A.2 Chapter 2

Exercise 2.5

(i)

The sequence (qn)n(q_{n})_{n\in\mathbb{N}} where qn0q_{n}\in\mathbb{N}_{0} is the unique value for which there exists some r{0,1,2,3,4}r\in\{0,1,2,3,4\} such that n=5qn+rn=5q_{n}+r.

(ii)

The sequence (rn)n(r_{n})_{n\in\mathbb{N}}, where rn{0,1,2,3,4}r_{n}\in\{0,1,2,3,4\} is the unique value for which there exists some q0q\in\mathbb{N}_{0} such that n=5q+rnn=5q+r_{n}.

(iii)

The sequence (rnqn+1)n\big{(}\frac{r_{n}}{q_{n}+1}\big{)}_{n\in\mathbb{N}} where qn0q_{n}\in\mathbb{N}_{0} and rn{0,1,2,3,4}r_{n}\in\{0,1,2,3,4\} are defined by n=5qn+rnn=5q_{n}+r_{n} for all nn\in\mathbb{N}.

Exercise 2.6

The sequence (an)n(a_{n})_{n\in\mathbb{N}} where an:=11/na_{n}:=1-1/n for nn even and an:=1/(n+1)a_{n}:=1/(n+1) for nn odd.

Exercise 2.8

(1(1, 11, 22, 33, 55, 88, 1313, 21,)21,\dots).

Exercise 2.9

The sequence is defined recursively by ‘saying out loud’ the digits of the previous term. In particular we have

1 is ‘one one’ so the next term is 1111,
11 is ‘two ones’ so the next term is 2121,
21 is ‘one two, one one’ so the next term is 12111211

and so on. Since 312211312211 is ‘one three, one one, two twos, two ones’, the next term of the sequence is 1311222113112221.

Exercise 2.13

Increasing Non-decreasing Decreasing Non-increasing Monotone
(i)
(ii)
(iii)
(iv)
(v)

Exercise 2.14

Suppose (an)n(a_{n})_{n\in\mathbb{N}} is a non-increasing sequence. Let nn\in\mathbb{N} so that, by hypothesis an+1ana_{n+1}\leq a_{n} and therefore an+1an-a_{n+1}\geq-a_{n}. Since this holds for all nn\in\mathbb{N}, we conclude that (an)n(-a_{n})_{n\in\mathbb{N}} is a non-decreasing sequence.

Exercise 2.18

(i) For all nn\in\mathbb{N}, we have

n+2n+2n=3nand2n2n2n2n2=n2,n+2\leq n+2n=3n\qquad\text{and}\qquad 2n^{2}-n\geq 2n^{2}-n^{2}=n^{2},

since n2nn^{2}\geq n for n1n\geq 1. Thus,

n+22n2n3nn2=3n3for all n,\frac{n+2}{2n^{2}-n}\leq\frac{3n}{n^{2}}=\frac{3}{n}\leq 3\qquad\text{for all % $n\in\mathbb{N}$,}

and therefore 33 is an upper bound for (n+22n2n)n\big{(}\tfrac{n+2}{2n^{2}-n}\big{)}_{n\in\mathbb{N}}. Since an upper bound exists, the sequence is bounded above.

(ii) For all nn\in\mathbb{N}, we have

10n4+n2+410n4and6n34n26n310n^{4}+n^{2}+4\geq 10n^{4}\qquad\text{and}\qquad 6n^{3}-4n^{2}\leq 6n^{3}

and therefore

an:=10n4+n2+46n34n210n46n3=10n6n.a_{n}:=\frac{10n^{4}+n^{2}+4}{6n^{3}-4n^{2}}\geq\frac{10n^{4}}{6n^{3}}=\frac{1% 0n}{6}\geq n.

Let MM\in\mathbb{R} and choose nn\in\mathbb{N} with n>Mn>M. The above inequality shows that an>Ma_{n}>M and so MM is not an upper bound for the sequence (an)n(a_{n})_{n\in\mathbb{N}}. Since MM was chosen arbitrarily, we conclude that this sequence is not bounded above.

Exercise 2.20

Suppose (an)n(a_{n})_{n\in\mathbb{N}} is bounded. Then there exists MM\in\mathbb{R} and LL\in\mathbb{R} such that LanML\leq a_{n}\leq M for all nn\in\mathbb{N}. If we choose R:=max{|L|,|M|}R:=\max\{|L|,|M|\}, then it follows that

R|L|LanM|M|R-R\leq-|L|\leq L\leq a_{n}\leq M\leq|M|\leq R

and therefore |an|R|a_{n}|\leq R for all nn\in\mathbb{N}.

Conversely, suppose there exists some R>0R>0 such that |an|R|a_{n}|\leq R for all nn\in\mathbb{N}. Taking M:=RM:=R and L:=RL:=-R, we have LanML\leq a_{n}\leq M for all nn\in\mathbb{N} and so (an)n(a_{n})_{n\in\mathbb{N}} is bounded.

Exercise 2.21

Suppose (an)n(a_{n})_{n\in\mathbb{N}} is bounded below. Then there exists some LL\in\mathbb{R} such that anLa_{n}\geq L for all nn\in\mathbb{N}. Defining M:=LM:=-L, it follows that anL=M-a_{n}\leq-L=M for all nn\in\mathbb{N} and so (an)n(-a_{n})_{n\in\mathbb{N}} is bounded above.

Exercise 2.28

(i) We claim that 3n2n+532\tfrac{3n}{2n+5}\to\frac{3}{2} as nn\to\infty.

For nn\in\mathbb{N}, we have

|3n2n+532|=|6n2(2n+5)3(2n+5)2(2n+5)|=52n+5<5n,\Big{|}\frac{3n}{2n+5}-\frac{3}{2}\Big{|}=\Big{|}\frac{6n}{2(2n+5)}-\frac{3(2n% +5)}{2(2n+5)}\Big{|}=\frac{5}{2n+5}<\frac{5}{n},

where in the last step we used the fact that 2n+5=n+n+5>n2n+5=n+n+5>n for all nn\in\mathbb{N}.

Let ε>0\varepsilon>0 be given and choose N:=5/εN:=\lceil 5/\varepsilon\rceil. If nn\in\mathbb{N} satisfies n>Nn>N, then it follows from our earlier work that

|3n2n+532|<5n<5Nε.\Big{|}\frac{3n}{2n+5}-\frac{3}{2}\Big{|}<\frac{5}{n}<\frac{5}{N}\leq\varepsilon.

Hence, by the ε\varepsilon-NN definition of a limit, 3n2n+532\tfrac{3n}{2n+5}\to\frac{3}{2} as nn\to\infty.

(ii) We claim that nn2+10\tfrac{n}{n^{2}+1}\to 0 as nn\to\infty.

For nn\in\mathbb{N}, we have

|nn2+10|=nn2+1<nn2=1n.\Big{|}\frac{n}{n^{2}+1}-0\Big{|}=\frac{n}{n^{2}+1}<\frac{n}{n^{2}}=\frac{1}{n}.

Let ε>0\varepsilon>0 be given and choose N:=1/εN:=\lceil 1/\varepsilon\rceil. If nn\in\mathbb{N} satisfies n>Nn>N, then

|nn2+10|<1n<1Nε.\Big{|}\frac{n}{n^{2}+1}-0\Big{|}<\frac{1}{n}<\frac{1}{N}\leq\varepsilon.

Hence, by the ε\varepsilon-NN definition of a limit, nn2+10\tfrac{n}{n^{2}+1}\to 0 as nn\to\infty.

Exercise 2.29

Let ε>0\varepsilon>0 be given and choose N:=1N:=1. Then for all nn\in\mathbb{N} with n>Nn>N, we have

|ana|=|aa|=0<ε.|a_{n}-a|=|a-a|=0<\varepsilon.

Thus, by the ε\varepsilon-NN definition of a limit, anaa_{n}\to a as nn\to\infty.

Exercise 2.31

Given 0<ρ<10<\rho<1, we may write

1ρ=(1ρ)(1+ρ)1+ρ=1ρ21+ρ.1-\rho=\frac{(1-\rho)(1+\rho)}{1+\rho}=\frac{1-\rho^{2}}{1+\rho}.

On the other hand, given nn\in\mathbb{N}, the Bernoulli inequality tells us that (1+ρ)n1+nρ(1+\rho)^{n}\geq 1+n\rho. Combining these observations,

(1ρ)n=(1ρ2)n(1+ρ)n11+nρ1nρ.(1-\rho)^{n}=\frac{(1-\rho^{2})^{n}}{(1+\rho)^{n}}\leq\frac{1}{1+n\rho}\leq% \frac{1}{n\rho}.

Now let 0<r<10<r<1 and write r=1ρr=1-\rho for some 0<ρ<10<\rho<1. Let ε>0\varepsilon>0 be given and choose N:=1/(ρε)N:=\lceil 1/(\rho\varepsilon)\rceil. If nn\in\mathbb{N} satisfies n>Nn>N, then

|rn0|=(1ρ)n1ρn<1ρNε.|r^{n}-0|=(1-\rho)^{n}\leq\frac{1}{\rho n}<\frac{1}{\rho N}\leq\varepsilon.

Thus, by the ε\varepsilon-NN definition of a limit, rn0r^{n}\to 0 as nn\to\infty.

Exercise 2.33

It is clear that a1=0=11=1100a_{1}=0=1-1=1-10^{0}. For nn\in\mathbb{N} with n2n\geq 2, we have

an=0.999n1-fold=10.00n2-fold1=110(n1).a_{n}=0.\underbrace{99\cdots 9}_{\text{$n-1$-fold}}=1-0.\underbrace{0\cdots 0}% _{\text{$n-2$-fold}}1=1-10^{-(n-1)}.

Hence, an=110(n1)a_{n}=1-10^{-(n-1)} for all nn\in\mathbb{N}.

Let ε>0\varepsilon>0 be given. From Exercise 2.31, we know that 10n010^{-n}\to 0 as nn\to\infty and so by the ε\varepsilon-NN definition of a limit there exists some NN\in\mathbb{N} such that 10n<ε/1010^{-n}<\varepsilon/10 for all nn\in\mathbb{N} with n>Nn>N. Consequently, if nn\in\mathbb{N} satisfies n>Nn>N, then

|an1|=10(n1)=1010n<ε.|a_{n}-1|=10^{-(n-1)}=10\cdot 10^{-n}<\varepsilon.

Hence, by the ε\varepsilon-NN definition of a limit, an1a_{n}\to 1 as nn\to\infty.

Exercise 2.35

For nn\in\mathbb{N}, we have

0n2+5n=(n2+5n)(n2+5+n)n2+5+n=5n2+5+n<5n.0\leq\sqrt{n^{2}+5}-n=\frac{(\sqrt{n^{2}+5}-n)(\sqrt{n^{2}+5}+n)}{\sqrt{n^{2}+% 5}+n}=\frac{5}{\sqrt{n^{2}+5}+n}<\frac{5}{n}.

Let ε>0\varepsilon>0 be given and choose N:=5/εN:=\lceil 5/\varepsilon\rceil\in\mathbb{N}. If nn\in\mathbb{N} with n>Nn>N, then

|n2+5n0|<5n<5Nε.|\sqrt{n^{2}+5}-n-0|<\frac{5}{n}<\frac{5}{N}\leq\varepsilon.

Thus, by the ε\varepsilon-NN definition of a limit, n2+5n0\sqrt{n^{2}+5}-n\to 0 as nn\to\infty.

Exercise 2.36

  1. (i)

    (a).

  2. (ii)

    (c).

  3. (iii)

    (b);

  4. (iv)

    (b);

  5. (v)

    (b);

  6. (vi)

    (d).

The justification is as follows:

(i) \iff (a). This is the ε\varepsilon-NN definition of a limit.

(ii) \iff (c). Suppose (ii) holds. Then taking N:=1N:=1, for all nn\in\mathbb{N} we have |ana|<ε|a_{n}-a|<\varepsilon for all ε>0\varepsilon>0. However, this can only hold if |ana|=0|a_{n}-a|=0 so that an=aa_{n}=a for all nn\in\mathbb{N}. Thus (c) holds.

Conversely, suppose (c) holds so that an=aa_{n}=a for all nn\in\mathbb{N}. Then for all ε>0\varepsilon>0 we have |ana|=0<ε|a_{n}-a|=0<\varepsilon for all nn\in\mathbb{N}, so (ii) holds.

(iii) \iff (b). Suppose (iii) holds. Taking N=1N=1, there exists some ε>0\varepsilon>0 such that |ana|<ε|a_{n}-a|<\varepsilon for all nn\in\mathbb{N}. Thus, by the triangle inequality, |an||a|+|ana|<|a|+ε|a_{n}|\leq|a|+|a_{n}-a|<|a|+\varepsilon for all nn\in\mathbb{N}. Thus, |an|M:=|a|+ε|a_{n}|\leq M:=|a|+\varepsilon for all nn\in\mathbb{N}, so (b) holds.

Conversely, suppose (b) holds. Then there exists some M>0M>0 such that |an|<M|a_{n}|<M for all nn\in\mathbb{N}. Let ε:=M+|a|\varepsilon:=M+|a|. Then |ana||an|+|a|<ε|a_{n}-a|\leq|a_{n}|+|a|<\varepsilon for all nn\in\mathbb{N}. Thus, (iii) holds.

(iv) \iff (b). Suppose (iv) holds, so there exists some NN\in\mathbb{N} and some ε>0\varepsilon>0 such that |ana|<ε|a_{n}-a|<\varepsilon for all n>Nn>N. Thus, by the triangle inequality, |an||a|+|ana|<|a|+ε|a_{n}|\leq|a|+|a_{n}-a|<|a|+\varepsilon for all n>Nn>N. From this we conclude that

|an|M:=max{|a1|,|a2|,,|aN|,|a|+ε}for all n,|a_{n}|\leq M:=\max\{|a_{1}|,|a_{2}|,\dots,|a_{N}|,|a|+\varepsilon\}\qquad% \text{for all $n\in\mathbb{N}$,}

so that (b) holds.

Conversely, suppose (b) holds. Then there exists some M>0M>0 such that |an|<M|a_{n}|<M for all nn\in\mathbb{N}. Let ε:=M+|a|\varepsilon:=M+|a|. Then |ana||an|+|a|<ε|a_{n}-a|\leq|a_{n}|+|a|<\varepsilon for all nn\in\mathbb{N}. Thus, (iii) holds.

(v) \iff (b). Suppose (v) holds. Taking N=1N=1, there exists some ε>0\varepsilon>0 such that |ana|<ε|a_{n}-a|<\varepsilon for all nn\in\mathbb{N}. Thus, |an|M:=|a|+ε|a_{n}|\leq M:=|a|+\varepsilon for all nn\in\mathbb{N}, so that (b) holds.

Conversely, suppose (b) holds. Then there exists some M>0M>0 such that |an|<M|a_{n}|<M for all nn\in\mathbb{N}. Let ε:=M+|a|\varepsilon:=M+|a|. Then |ana||an|+|a|<ε|a_{n}-a|\leq|a_{n}|+|a|<\varepsilon for all nn\in\mathbb{N}. Thus, (v) holds.

(vi) \iff (d). The condition |ana|<ε|a_{n}-a|<\varepsilon for all ε>0\varepsilon>0 holds if and only if |ana|=0|a_{n}-a|=0, which in turn holds if and only if an=aa_{n}=a. Thus, (vi) holds if and only if there exists some NN\in\mathbb{N} such that an=aa_{n}=a for all n>Nn>N, which is precisely the condition (d).

Exercise 2.40

(i) We claim that 10n3+5n+1n3+3n2+510\frac{10n^{3}+5n+1}{n^{3}+3n^{2}+5}\to 10 as nn\to\infty.

Multiplying both the numerator and denominator by 1/n31/n^{3} gives

10n3+5n+1n3+3n2+5=10+5/n2+1/n31+3/n+5/n3.\frac{10n^{3}+5n+1}{n^{3}+3n^{2}+5}=\frac{10+5/n^{2}+1/n^{3}}{1+3/n+5/n^{3}}.

Since we know 1/n01/n\to 0 as nn\to\infty, it follows from parts 1 and 2 of Theorem 2.38 that

limn10+5/n2+1/n3=10andlimn1+3/n+5/n3=1.\lim_{n\to\infty}10+5/n^{2}+1/n^{3}=10\qquad\text{and}\qquad\lim_{n\to\infty}1% +3/n+5/n^{3}=1.

Since the limit of the denominator is non-zero, we can apply part 3 of Theorem 2.38 to conclude that

limn10n3+5n+1n3+3n2+5=101=10,\lim_{n\to\infty}\frac{10n^{3}+5n+1}{n^{3}+3n^{2}+5}=\frac{10}{1}=10,

as required.

(ii) We claim that 2n+32n+1212\tfrac{2^{n}+3}{2^{n+1}-2}\to\tfrac{1}{2} as nn\to\infty.

Multiplying both the numerator and denominator by 2n2^{-n} gives

2n+32n+12=1+32n222n.\frac{2^{n}+3}{2^{n+1}-2}=\frac{1+3\cdot 2^{-n}}{2-2\cdot 2^{-n}}.

Since we know from Exercise 2.31 that 2n02^{-n}\to 0 as nn\to\infty, it follows from parts 1 and 2 of Theorem 2.38 that

limn1+32n=1andlimn222n=2.\lim_{n\to\infty}1+3\cdot 2^{-n}=1\qquad\text{and}\qquad\lim_{n\to\infty}2-2% \cdot 2^{-n}=2.

Since the limit of the denominator is non-zero, we can apply part 3 of Theorem 2.38 to conclude that

limn2n+32n+12=12,\lim_{n\to\infty}\frac{2^{n}+3}{2^{n+1}-2}=\frac{1}{2},

as required.

Exercise 2.41

For nn\in\mathbb{N}, observe that

|1bn1b|=|bnb||b||bn|.\Big{|}\frac{1}{b_{n}}-\frac{1}{b}\Big{|}=\frac{|b_{n}-b|}{|b||b_{n}|}.

Since bnbb_{n}\to b as nn\to\infty, we expect to be able to replace the bnb_{n} on the denominator with bb. More precisely, by the ε\varepsilon-NN definition of a limit, since b0b\neq 0 there exists some N1N_{1}\in\mathbb{N} such that

|bnb|<|b|2for all n with n>N1.|b_{n}-b|<\frac{|b|}{2}\qquad\text{for all $n\in\mathbb{N}$ with $n>N_{1}$.}

If nn\in\mathbb{N} satisfies n>N1n>N_{1}, then it follows by the triangle inequality that

|bn||b||bnb|>|b||b|/2=|b|/2|b_{n}|\geq|b|-|b_{n}-b|>|b|-|b|/2=|b|/2

and so

|1bn1b|=|bnb||b||bn|<2|bnb||b|2.\Big{|}\frac{1}{b_{n}}-\frac{1}{b}\Big{|}=\frac{|b_{n}-b|}{|b||b_{n}|}<\frac{2% |b_{n}-b|}{|b|^{2}}.

Let ε>0\varepsilon>0. Since bnbb_{n}\to b as nn\to\infty and b0b\neq 0, by the ε\varepsilon-NN definition of a limit, there exists some N2N_{2}\in\mathbb{N} such that

|bnb|<|b|2ε2for all n with n>N2.|b_{n}-b|<\frac{|b|^{2}\varepsilon}{2}\qquad\text{for all $n\in\mathbb{N}$ % with $n>N_{2}$.}

If we define N:=max{N1,N2}N:=\max\{N_{1},N_{2}\}, then

|bnb|<min{|b|2,|b|2ε2}for all n with n>N.|b_{n}-b|<\min\Big{\{}\frac{|b|}{2},\frac{|b|^{2}\varepsilon}{2}\Big{\}}\qquad% \text{for all $n\in\mathbb{N}$ with $n>N$.}

If nn\in\mathbb{N} satisfies n>N1n>N_{1}, then it follows from our earlier observations that

|1bn1b|=|bnb||b||bn|2|bnb||b|2<2|b|2|b|2ε2=ε.\Big{|}\frac{1}{b_{n}}-\frac{1}{b}\Big{|}=\frac{|b_{n}-b|}{|b||b_{n}|}\leq% \frac{2|b_{n}-b|}{|b|^{2}}<\frac{2}{|b|^{2}}\cdot\frac{|b|^{2}\varepsilon}{2}=\varepsilon.

Hence, by the ε\varepsilon-NN definition of a limit, 1/bn1/b1/b_{n}\to 1/b as nn\to\infty.

Exercise 2.44

For nn\in\mathbb{N}, we have

2n12n+5(1)n2n+512n+52n.-2^{-n}\leq\frac{-1}{2^{n}+5}\leq\frac{(-1)^{n}}{2^{n}+5}\leq\frac{1}{2^{n}+5}% \leq 2^{-n}.

We know from Exercise 2.31 that 2n02^{-n}\to 0 as nn\to\infty and, by the limit laws, 2n0-2^{-n}\to 0 as nn\to\infty. Hence, by the squeeze theorem, (1)n2n+50\frac{(-1)^{n}}{2^{n}+5}\to 0 as nn\to\infty.

Exercise 2.45

Let AA\subseteq\mathbb{R} be a nonempty set which is bounded above, so that s:=supAs:=\sup A exists by the completeness axiom.

(i) Given nn\in\mathbb{N}, using the approximation property for suprema from Lemma 1.31 with ε:=1/n\varepsilon:=1/n, there exists some anAa_{n}\in A such that s1/n<anss-1/n<a_{n}\leq s.

(ii) From part (i) we have nanun\ell_{n}\leq a_{n}\leq u_{n} where n:=s1/n\ell_{n}:=s-1/n and un:=su_{n}:=s for all nn\in\mathbb{N}. Since ns\ell_{n}\to s and unsu_{n}\to s as nn\to\infty, it follows from the squeeze theorem that ansa_{n}\to s as nn\to\infty.

Exercise 2.46

(i) We argue by contradiction. Suppose (an)n(a_{n})_{n\in\mathbb{N}} and (bn)n(b_{n})_{n\in\mathbb{N}} are convergent sequences with anbna_{n}\leq b_{n} for all nn\in\mathbb{N} and

a:=limnanandb:=limnbnsatisfya>b.a:=\lim_{n\to\infty}a_{n}\quad\text{and}\quad b:=\lim_{n\to\infty}b_{n}\qquad% \text{satisfy}\qquad a>b.

Let ε:=(ab)/2>0\varepsilon:=(a-b)/2>0. Then by the ε\varepsilon-NN definition of a limit there exists some N1N_{1}\in\mathbb{N} such that |aan|<ε|a-a_{n}|<\varepsilon for all n>N1n>N_{1}. Similarly, there exists some N2N_{2}\in\mathbb{N} such that |bbn|<ε|b-b_{n}|<\varepsilon for all n>N2n>N_{2}. Taking N:=max{N1,N2}N:=\max\{N_{1},N_{2}\}, it follows that |aan|<ε|a-a_{n}|<\varepsilon and |bbn|<ε|b-b_{n}|<\varepsilon for all n>Nn>N. Consequently, for n>Nn>N we have

a+b2=aab2=aε<anbn<b+ε=b+ab2=a+b2\frac{a+b}{2}=a-\frac{a-b}{2}=a-\varepsilon<a_{n}\leq b_{n}<b+\varepsilon=b+% \frac{a-b}{2}=\frac{a+b}{2}

and so

a+b2<a+b2.\frac{a+b}{2}<\frac{a+b}{2}.

This is a contradiction, so we must have aba\leq b.

(ii) This is false. Indeed, consider the convergent sequences (an)n(a_{n})_{n\in\mathbb{N}} and (bn)n(b_{n})_{n\in\mathbb{N}} given by an:=0a_{n}:=0 and bn:=1/nb_{n}:=1/n for all nn\in\mathbb{N}. Then an<bna_{n}<b_{n} for all nn\in\mathbb{N} but limnan=limnbn=0\displaystyle\lim_{n\to\infty}a_{n}=\lim_{n\to\infty}b_{n}=0.

Exercise 2.52

The proof of the boundedness test. If anaa_{n}\to a as nn\to\infty, then there exists some NN\in\mathbb{N} such that |an||a|+1|a_{n}|\leq|a|+1 for all n>Nn>N. We can then take a maximum of the set of remaining terms to conclude that |an|M:=max{|a1|,,|aN1|,|a|+1}|a_{n}|\leq M:=\max\{|a_{1}|,\dots,|a_{N-1}|,|a|+1\} for all nn\in\mathbb{N}.

Exercise 2.56

(i) For nn\in\mathbb{N}, we have

n3n+1n32n=n22n2.\frac{n^{3}}{n+1}\geq\frac{n^{3}}{2n}=\frac{n^{2}}{2}\geq\frac{n}{2}.

Let M>0M>0 be given and choose N:=2MN:=\lceil 2M\rceil. If nn\in\mathbb{N} satisfies n>Nn>N, then

n3n+1n2>2M2=M.\frac{n^{3}}{n+1}\geq\frac{n}{2}>\frac{2M}{2}=M.

Thus, by definition, n3n+1\frac{n^{3}}{n+1}\to\infty as nn\to\infty.

(ii) From Workshop 3, we know n2n0n2^{-n}\to 0 as nn\to\infty. Hence, given M>0M>0, taking ε:=M1>0\varepsilon:=M^{-1}>0 in the ε\varepsilon-NN definition of a limit, there exists some NN\in\mathbb{N} such that 0<n2n<M10<n2^{-n}<M^{-1} for all n>Nn>N. Taking reciprocals, it follows that if n>Nn>N, then 2n/n>M2^{n}/n>M. Thus, by definition, 2n/n2^{n}/n\to\infty as nn\to\infty.

(iii) For nn\in\mathbb{N}, by dividing through by 22n=4n2^{2n}=4^{n}, we have

22n3n+n2=1(3/4)n+n24n.\frac{2^{2n}}{3^{n}+n^{2}}=\frac{1}{(3/4)^{n}+n^{2}4^{-n}}.

By Exercise 2.31 and Worksheet 3, we know (3/4)n0(3/4)^{n}\to 0 and n24n0n^{2}4^{-n}\to 0 as nn\to\infty. By the limit laws, (3/4)n+n24n0(3/4)^{n}+n^{2}4^{-n}\to 0 as nn\to\infty. Hence, given M>0M>0, taking ε:=M1>0\varepsilon:=M^{-1}>0 in the ε\varepsilon-NN definition of a limit, there exists some NN\in\mathbb{N} such that 0<(3/4)n+n24n<M10<(3/4)^{n}+n^{2}4^{-n}<M^{-1} for all n>Nn>N. Taking reciprocals, it follows that if n>Nn>N, then 22n3n+n2>M\tfrac{2^{2n}}{3^{n}+n^{2}}>M. Thus, by definition, 22n3n+n2\tfrac{2^{2n}}{3^{n}+n^{2}}\to\infty as nn\to\infty

Exercise 2.57

Let an:=(1)nna_{n}:=(-1)^{n}n for all nn\in\mathbb{N}. Then (an)n(a_{n})_{n\in\mathbb{N}} is unbounded. Indeed, given MM\in\mathbb{R}, if we choose n>|M|n>|M| odd, then it follows that an=n<|M|Ma_{n}=-n<-|M|\leq M. On the other hand, if we chose n>|M|n>|M| even, then an=n>|M|Ma_{n}=n>|M|\geq M. Thus, MM is neither a lower or an upper bound for (an)n(a_{n})_{n\in\mathbb{N}}. Since MM\in\mathbb{R} was chosen arbitrarily, it follows that (an)n(a_{n})_{n\in\mathbb{N}} is not bounded above or below.

We claim that an↛a_{n}\not\to\infty as nn\to\infty. Indeed, if ana_{n}\to\infty as nn\to\infty, then there would exist some NN\in\mathbb{N} such that an>0a_{n}>0 for all n>Nn>N. However, a2n+1=(2n+1)<0a_{2n+1}=-(2n+1)<0 for all nn\in\mathbb{N}, so there can be no such choice of NN.

Finally, we claim that an↛a_{n}\not\to-\infty as nn\to\infty. Indeed, if ana_{n}\to-\infty as nn\to\infty, then there would exist some NN\in\mathbb{N} such that an<0a_{n}<0 for all n>Nn>N. However, a2n=2n>0a_{2n}=2n>0 for all nn\in\mathbb{N}, so there can be no such choice of NN.

Exercise 2.62

Suppose (an)n(a_{n})_{n\in\mathbb{N}} is sequence which converges to a limit aa\in\mathbb{R} and (ank)k(a_{n_{k}})_{k\in\mathbb{N}} is a subsequence. Let ε>0\varepsilon>0 be given. By applying the ε\varepsilon-NN definition of a limit to (an)n(a_{n})_{n\in\mathbb{N}}, there exists some NN\in\mathbb{N} such that |ana|<ε|a_{n}-a|<\varepsilon for all n>Nn>N. However, if k>Nk>N, then it follows that nk>k>Nn_{k}>k>N and so |anka|<ε|a_{n_{k}}-a|<\varepsilon. Thus, by the ε\varepsilon-NN definition of a limit, ankaa_{n_{k}}\to a as kk\to\infty.

Exercise 2.65

Let an:=sin2(πn50)a_{n}:=\sin^{2}\big{(}\frac{\pi n}{50}\big{)} for nn\in\mathbb{N}. Then

a50k=sin2(πk)=0for all ka_{50k}=\sin^{2}(\pi k)=0\qquad\text{for all $k\in\mathbb{N}$}

and

a50k+25=sin2(πk+π2)=1for all k.a_{50k+25}=\sin^{2}\Big{(}\pi k+\frac{\pi}{2}\Big{)}=1\qquad\text{for all $k% \in\mathbb{N}$.}

Hence, a50k0a_{50k}\to 0 as kk\to\infty and a50k+251a_{50k+25}\to 1 as kk\to\infty. In particular, we have found two subsequences with different limits, and so the sequence (an)n(a_{n})_{n\in\mathbb{N}} must diverge by the subsequence test.

Exercise 2.66

We show (qn)n(q_{n})_{n\in\mathbb{N}} diverges by the boundedness test. Indeed, let M>0M>0 and choose n>5M+5n>5M+5. Then qnq_{n} satisfies n=5qn+rnn=5q_{n}+r_{n} where 0rn40\leq r_{n}\leq 4 and so

(A.3) (A.3) qn=nrn5n45>M.q_{n}=\frac{n-r_{n}}{5}\geq\frac{n-4}{5}>M.

Thus, MM is not an upper bound for (qn)n(q_{n})_{n\in\mathbb{N}}. Since M>0M>0 was chosen arbitrarily, it follows that (qn)n(q_{n})_{n\in\mathbb{N}} is unbounded and therefore diverges by the boundedness test.

We show (rn)n(r_{n})_{n\in\mathbb{N}} diverges by the subsequence test. Indeed, for all kk\in\mathbb{N} we have r5k=0r_{5k}=0 and r5k+1=1r_{5k+1}=1. Thus, r5k0r_{5k}\to 0 and r5k+11r_{5k+1}\to 1 as kk\to\infty. Hence, (rn)n(r_{n})_{n\in\mathbb{N}} has two distinct subsequential limits and therefore diverges by the subsequence test.

Finally, we claim rnqn+10\tfrac{r_{n}}{q_{n}+1}\to 0 as nn\to\infty. Indeed, let ε>0\varepsilon>0 and choose N:=20/ε+4N:=\lceil 20/\varepsilon\rceil+4. If n>Nn>N, then arguing as in (A.3) we see that

qnn45>N454ε.q_{n}\geq\frac{n-4}{5}>\frac{N-4}{5}\geq\frac{4}{\varepsilon}.

Recalling that 0rn40\leq r_{n}\leq 4 for all nn\in\mathbb{N}, we therefore have

|rnqn+10|=rnqn+14qn<ε.\Big{|}\frac{r_{n}}{q_{n}+1}-0\Big{|}=\frac{r_{n}}{q_{n}+1}\leq\frac{4}{q_{n}}% <\varepsilon.

Thus, by the ε\varepsilon-NN definition of a limit, rnqn+10\tfrac{r_{n}}{q_{n}+1}\to 0 as nn\to\infty.

Exercise 2.67

(i) Let aa\in\mathbb{R} and ε:=1\varepsilon:=1. Given NN\in\mathbb{N}, choose n>max{N,|a|+1}n>\max\{N,|a|+1\}, so that n>Nn>N. It follows that |na||n||a|>1=ε|n-a|\geq|n|-|a|>1=\varepsilon.

The above argument shows that for all aa\in\mathbb{R}, there exists some ε>0\varepsilon>0 such that for all NN\in\mathbb{N} there exists some n>Nn>N such that |na|>ε|n-a|>\varepsilon. Hence, by the ε\varepsilon-NN definition, (n)n(n)_{n\in\mathbb{N}} diverges.

(ii) Let aa\in\mathbb{R} and ε:=1/2\varepsilon:=1/2. Let σ{0,1}\sigma\in\{0,1\} satisfy

|(1)σa|=maxj{0,1}|(1)ja|;|(-1)^{\sigma}-a|=\max_{j\in\{0,1\}}|(-1)^{j}-a|;

in other words, σ:=0\sigma:=0 if |1a||1+a||1-a|\geq|1+a| and σ:=1\sigma:=1 if |1+a||1a||1+a|\geq|1-a|. It then follows from the definition that

2|(1)σa|=|(1)σa|+|(1)σa||(1)0a|+|(1)1a|.2|(-1)^{\sigma}-a|=|(-1)^{\sigma}-a|+|(-1)^{\sigma}-a|\geq|(-1)^{0}-a|+|(-1)^{% 1}-a|.

Furthermore, by the triangle inequality,

2|(1)σa||1a|+|(1)a||1a(1)+a|=22|(-1)^{\sigma}-a|\geq|1-a|+|(-1)-a|\geq|1-a-(-1)+a|=2

and so |(1)σa|1|(-1)^{\sigma}-a|\geq 1.

Given NN\in\mathbb{N}, choose n>Nn>N even if σ=0\sigma=0 and n>Nn>N odd if σ=1\sigma=1, so that (1)n=(1)σ(-1)^{n}=(-1)^{\sigma}. Thus,

|(1)na|=|(1)σa|1>1/2=ε.|(-1)^{n}-a|=|(-1)^{\sigma}-a|\geq 1>1/2=\varepsilon.

The above argument shows that for all aa\in\mathbb{R}, there exists some ε>0\varepsilon>0 such that for all NN\in\mathbb{N} there exists some n>Nn>N such that |(1)na|>ε|(-1)^{n}-a|>\varepsilon. Hence, by the ε\varepsilon-NN definition, ((1)n)n((-1)^{n})_{n\in\mathbb{N}} diverges.

Exercise 2.70

Let (an)n(a_{n})_{n\in\mathbb{N}} be non-increasing and bounded below. By Exercise 2.14 and Exercise 2.21, the sequence (an)n(-a_{n})_{n\in\mathbb{N}} is non-decreasing and bounded above. Hence, by part 1 of the monotone convergence theorem, (an)n(-a_{n})_{n\in\mathbb{N}} converges. Finally, by the limit laws, (an)n(a_{n})_{n\in\mathbb{N}} = ((an))n(-(-a_{n}))_{n\in\mathbb{N}} converges.

Exercise 2.71

(i) We prove an>0a_{n}>0 for all nn\in\mathbb{N} by induction on nn.

Base case: By hypothesis, a1=2>0a_{1}=2>0, which forms the base case of our induction.

Inductive step: Suppose, as an induction hypothesis, that an>0a_{n}>0 for some nn\in\mathbb{N}. Then using the formula

an+1=12(an+2an),a_{n+1}=\frac{1}{2}\Big{(}a_{n}+\frac{2}{a_{n}}\Big{)},

we see that an+1a_{n+1} is formed by multiplying together and summing positive terms and therefore must itself be positive. This closes the induction.

(ii) Let nn\in\mathbb{N}. Using the formula

an+1=an12(an2an),a_{n+1}=a_{n}-\frac{1}{2}\Big{(}a_{n}-\frac{2}{a_{n}}\Big{)},

we expand out the square to give

an+12=an2an(an2an)+14(an2an)2=2+14(an2an)2,a_{n+1}^{2}=a_{n}^{2}-a_{n}\Big{(}a_{n}-\frac{2}{a_{n}}\Big{)}+\frac{1}{4}\Big% {(}a_{n}-\frac{2}{a_{n}}\Big{)}^{2}=2+\frac{1}{4}\Big{(}a_{n}-\frac{2}{a_{n}}% \Big{)}^{2},

as required. Since the squared term is always non-negative, we conclude that an+122a_{n+1}^{2}\geq 2 for all nn\in\mathbb{N}. Thus, since a122a_{1}^{2}\geq 2 by hypothesis, we have an22a_{n}^{2}\geq 2 for all nn\in\mathbb{N}.

(iii) From part (i) we know the sequence (an)n(a_{n})_{n\in\mathbb{N}} is bounded below by 0. We claim (an)n(a_{n})_{n\in\mathbb{N}} is monotone non-increasing. To see this, let nn\in\mathbb{N} and consider

(A.4) (A.4) anan+1=12(an2an)=12(an22an).a_{n}-a_{n+1}=\frac{1}{2}\Big{(}a_{n}-\frac{2}{a_{n}}\Big{)}=\frac{1}{2}\Big{(% }\frac{a_{n}^{2}-2}{a_{n}}\Big{)}.

By part (i), we know an>0a_{n}>0 and by part (ii) we know an220a_{n}^{2}-2\geq 0. Hence the right-hand side of (A.4) is non-negative and so anan+1a_{n}\geq a_{n+1}. Thus the sequence is monotone non-increasing.

(iv) By part (iii) and the monotone convergence theorem for sequences, (an)n(a_{n})_{n\in\mathbb{N}} is convergent. Let aa\in\mathbb{R} denote the limit of the sequence. By taking the limit of both sides of the relation

an+1=12(an+2an),a_{n+1}=\frac{1}{2}\Big{(}a_{n}+\frac{2}{a_{n}}\Big{)},

we see that aa must satisfy

a=12(a+2a).a=\frac{1}{2}\Big{(}a+\frac{2}{a}\Big{)}.

Rearranging this identity, we conclude that a2=2a^{2}=2, as required.

Exercise 2.72

The sequence (an)n=(0,0.3,0.33,0.333,)(a_{n})_{n\in\mathbb{N}}=(0,0.3,0.33,0.333,\dots) is clearly increasing (and therefore monotone) and bounded above by 11. Thus, by the monotone convergence theorem, anaa_{n}\to a as nn\to\infty for some aa\in\mathbb{R}.

By the limit laws, we know 3an3a3a_{n}\to 3a as nn\to\infty, where the sequence (3an)n(3a_{n})_{n\in\mathbb{N}} is given by (0,0.9,0.99,0.999,)(0,0.9,0.99,0.999,\dots). However, we know from Exercise 2.33 that this converges to 11 and so by the uniqueness of limits, 3a=13a=1.

Exercise 2.73

(i) When 0<r10<r\leq 1 the sequence r1/nr^{1/n} is monotone non-decreasing and bounded above by 11; for r>1r>1 the sequence r1/nr^{1/n} is monotone decreasing bounded below by 11. In either case, by the monotone convergence theorem, the sequence must converge to some limit LL\in\mathbb{R}.

In the case r>1r>1, we know the sequence (rn)n(r^{n})_{n\in\mathbb{N}} is bounded below by 11. Thus, we must have L1L\geq 1 and, in particular, L>0L>0. On the other hand, for 0<r10<r\leq 1 we know the sequence is monotone non-decreasing, and so we must have La1=r>0L\geq a_{1}=r>0.

(ii) Let an:=rna_{n}:=r^{n} for all nn\in\mathbb{N} and consider the squared sequence (an2)n(a_{n}^{2})_{n\in\mathbb{N}}. By the product law for limits, this sequence must converge to L2L^{2}. Now consider the subsequence of even terms (a2k2)k(a_{2k}^{2})_{k\in\mathbb{N}}, given by

a2k2=(a1/2k)2=a1/k=akfor all k.a_{2k}^{2}=(a^{1/2k})^{2}=a^{1/k}=a_{k}\qquad\text{for all $k\in\mathbb{N}$.}

Consequently, the subsequential limit is given by

limka2k2=limkak=L.\lim_{k\to\infty}a_{2k}^{2}=\lim_{k\to\infty}a_{k}=L.

However, by the subsequence test, the subsequential limit must equal the limit of the parent sequence, and so L2=LL^{2}=L.

Since L2=LL^{2}=L, or in other words L(L1)=L2L=0L(L-1)=L^{2}-L=0, we must have either L=1L=1 or L=0L=0. However, we showed in part (i) that L>0L>0 and so we must have L=1L=1.

Exercise 2.75

If we take the unbounded sequence (an)n(a_{n})_{n\in\mathbb{N}} given by an:=na_{n}:=n for all nn\in\mathbb{N}, then this has no convergent subsequence. Indeed, given any strictly increasing sequence (nk)k(n_{k})_{k\in\mathbb{N}} of natural numbers, we have nkkn_{k}\geq k for all kk\in\mathbb{N} and so (ank)k=(nk)k(a_{n_{k}})_{k\in\mathbb{N}}=(n_{k})_{k\in\mathbb{N}} is unbounded and therefore diverges by the boundedness test.