Solutions to Chapter 3
Exercise 3.6¶
Consider the sequence
\[ \frac{1}{2},\frac{1}{3},\frac{2}{3},\frac{1}{4},\frac{2}{4},\frac{3}{4}, \cdots \]For which numbers \(L\) does the above have a subsequence converging to \(L\)? Furthermore, does there exist a sequence \((a_n)\) where for every \(L\in \mathbb{R}\) there is a subsequence of \((a_n)\) which converges to \(L\)?
Answer
The second part has already been answered here. After reading the solution, the first part should be obvious.
Exercise 3.8¶
Assume that \((a_n)\) converges to some \(a\in \mathbb{R}\) and \((b_n)\) converges to some \(b\in \mathbb{R}\). Also assume \(c\in \mathbb{R}\).
- Prove that \((a_n+b_n)\) converges to \(a+b\)
- Prove that \((c\cdot a_n)\) converges to \(ca\)
Answer
Recall that we proved \(|ab| =|a||b|\) for any real \(a,b\) in Exercise 1.14. We will be relying on this without reference moving forward.
Part 1. Let \(\epsilon>0\), then there exists \(N_1, N_2\) such that \(|a_n-a|<\epsilon/2\) whenever \(n> N_1\) and \(|b_n-b|<\epsilon/2\) whenever \(n> N_2\). Let \(N:= \max\{N_1,N_2\}\), thus \(|a_n-a| + |b_n-b| < \epsilon\) whenever \(n\geq N\), and by triangle inequality :
whenever \(n\geq N\) as needed.
Part 2. The case when \(c=0\) is obvious, so assume \(c\neq 0\) and let \(\epsilon>0\), then there exists an \(N\) such that \(|a_n-a|<\frac{\epsilon}{|c|}\) whenever \(n\geq N\). Multiplying by \(|c|\) on both sides gives
whenever \(n\geq N\) as needed.
Exercise 3.10¶
Assuming that \((a_{2n})\) converges to \(L\) and \((a_{2n-1})\) converges to \(L\), show that \((a_n)\) also converges to \(L\).
Answer
I answered this in "An "Almost Always" Approach to Sequences
Let \(\epsilon>0\). Then there is \(N_1\) and \(N_2\) such that \(|a_{2n}-L|<\epsilon\) whenever \(n>N_1\) and \(|a_{2n-1}-L|<\epsilon\) whenever \(n>N_2\). Let \(N:=\max\{2N_1,2N_2-1\}\) and assume \(m>N\). If \(m\) is even then \(m=2n\) for some \(n\) which implies \(n>N_1\) and hence \(|a_m-L|=|a_{2n}-L|<\epsilon\). If \(m\) is odd then \(m=2n-1\) which implies \(n>N_2\) and hence \(|a_{2n-1}-L|=|a_m-L|<\epsilon\). Thus \((a_n)\to L\) as needed.
Exercise 3.12¶
Suppose \((a_n)\) is a sequence and \(f:\mathbb{N}\to \mathbb{N}\) is a bijection. For each of the following, prove that the statement is true or find a counterexample showing that the statement is false.
- If \((a_n)\) diverges to \(\infty\), then \((a_{f(n)})\) diverges to \(\infty\)
- If \((a_n)\) converges to \(L\), then \((a_{f(n)})\) converges to \(L\).
- If \((a_n)\)'s limit does not exist, then \((a_{f(n)})\)'s limit does not exist.
Answer
I will just prove the first two parts for which the property that \(f\) is injective is sufficient so we shall assume that that is the case moving forward. We shall prove two lemmas, from which parts 1 and 2 should be immediately obvious
Lemma 1. Suppose \((a_n)\to L\) where \(L\) can be real or \(\pm\infty\). For any sequence \((k_n)\) of natural numbers that diverges to \(\infty\), we have \((a_{n_k})\to L\).
Proof. We prove the case for when \(L\in \mathbb{R}\), the proof is similar for \(L=\pm \infty\). Let \(\epsilon>0\), then there is an \(N\) such that \(|a_n-L|<\epsilon\) whenever \(n>N\). Since \((k_n)\to \infty\) there is an \(N'\) such that \(k_n>N\) whenever \(n>N'\). The idea is that \(n>N' \ \Rightarrow \ k_n>N \ \Rightarrow \ |a_{k_n}-L|<\epsilon\), thus \((a_{n_k})\to L\) as needed.
Lemma 2. Let \(f:\mathbb{N}\to \mathbb{N}\) be injective, then \(f(n)\to \infty\), in other words : for any \(M\in \mathbb{R}\) there exists an \(N\in \mathbb{N}\) such that \(f(n)>M\) whenever \(n>N\).
Proof. Set \(A:=\{1,2,\ldots,M\}\) and consider \(f^{-1}(A)\). The inverse image of a finite set is finite so \(f^{-1}(A)\) is finite. This means that there is an \(N\) large enough such that
\[f^{-1}(A)\subseteq \{n\in \mathbb{N} \ \vert \ n\leq N\} \quad \Longrightarrow \quad \{n\in \mathbb{N} \ \vert \ n\leq N\}^C \subseteq f^{-1}(A)^C\]Thus, \(n\in f^{-1}(A)^C \ \Rightarrow \ f(n)\notin A \ \Rightarrow \ f(n) > M\) whenever \(n>N\) as required.
Exercise 3.18¶
Assume that \((a_n)\) is bounded and \((b_n)\) converges to 0. Prove that \((a_n\cdot b_n)\) also converges to 0.
Give an example of sequences \((a_n)\) and \((b_n)\) such that \(b_n\to 0\) but \((a_n\cdot b_n)\) does not converge to zero.
Answer
Part 1. Since \((a_n)\) is bounded, there exists some positive \(C\in \mathbb{R}\) such that \(|a_n|<C\) for all \(n\). Let \(\epsilon>0\). Since \((b_n)\to 0\) there exists an \(N\) such that \(|b_n|<\epsilon / C\) whenever \(n>N\). Combining both inequalities, we get :
whenever \(n>N\), thus \((a_n\cdot b_n)\to 0\) as needed.
Part 2. Consider \((a_n) = (1,2,3,\ldots)\) and \(b_n = (1, 1/2, 1/3, \ldots)\). Then clearly \((b_n)\to 0\) but \((a_n\cdot b_n) = (1,1,1,\ldots)\to 1\).
Exercise 3.23¶
- Prove that if \(a_n\to L\), then \(|a_n|\to |L|\)
- Give an example where \(|a_n|\to |L|\) but \(a_n\not\to L\)
Answer
Part 1. By the reverse triangle inequality ||a_n|-|L||\leq |a_n-L|$ for all \(n\). Thus, our task is equivalent to showing that for any \(\epsilon>0\) there is an \(N\) such that \(|a_n-L|<\epsilon\) whenever \(n>N\), but this is precisely what \(a_n\to L\) means.
Part 2. Let \((a_n):=(1,-1,1,-1,\ldots)\) and \(L:=1\). Then clearly \(|a_n|\to |L|\) but \(a_n \not\to L\).
Exercise 3.28¶
Prove that the sequence \((r^n)\) converges to 0 if \(r\in(-1,1)\), converges to \(1\) if \(r=1\), and diverges otherwise.
Lemma 1
Lemma 1. (Bernoulli's Inequality). For any real \(x>-1\) and integer \(n\geq 1\) :
\[ (1+x)^n \geq 1+nx \]
Proof. It is easy to show by induction.
Lemma 2
Lemma 2. \(a_n\to 0 \ \Leftrightarrow \ |a_n|\to 0\).
Proof. The forward direction has already been shown by Exercise 3.23. The reverse direction is very easy to show.
Answer
We split this into 6 cases, the first three are obvious :
- Convergence to 1 when \(r=1\)
- Convergence to 0 when \(r=0\)
- Divergence when \(r=-1\)
- Convergence to 0 when \(0<|r|<1\)
- \(r^n\) diverges when \(r>1\)
- \(r^n\) diverges when \(r<-1\)
Case 4. \(r^n\to 0\) whenever \(0<|r|<1\)
By Lemma 2, we can prove \(|r^n|\to 0\) instead of \(r^n\to 0\). Since \(0<|r|<1\), taking the reciprocal we have \(1<1/|r|\). This means we can write \(1/|r|=1+h\) for some \(h>0\). We use Bernoulli's Inequality with \(x:=h\) to get
for any integer \(n\geq 1\). Now 0 is a lower bound for \(|r^n|\) and clearly \(\frac{1}{1+nh}\to 0\) as \(n\to \infty\), thus \(|r^n|\to 0\) by the squeeze theorem.
Case 5. \(r^n\) diverges when \(r>1\)
We can write \(r=1+h\) for some \(h>0\). We use Bernoulli's inequality with \(x:=h\) to get
The RHS grows without bound as \(n\to \infty\) thus it is clear that \(r^n\to \infty\) as well.
Case 6. \(r^n\) diverges when \(r<1\)
By the contrapositive of Exercise 3.23 : If \(|r^n|\) diverges, then so does \(r^n\). If \(r<-1\) then \(|r|>1\) so \(|r|^n\) diverges by Case 5 and we are done.
Exercise 3.29¶
Suppose \((a_n)\) is a sequence for which \(a_n\to a\). Define a new sequence by
\[ b_n := \frac{a_1+a_2+\cdots + a_n}{n} \]Prove that \(b_n\to a\).
Answer
Exercise 3.30¶
Exercise 3.33¶
Exercise 3.36¶
Exercise 3.39¶
Exercise 3.41¶
Exercise 3.43¶
Exercise 3.44¶
Misc 1¶
Suppose \((a_n)\) is a sequence and \(a\in \mathbb{R}\). Prove that \(a_n\to a\) iff \(|a_n-a|\to 0\)
Answer
\(|a_n-a|\to 0\) means that for any \(\epsilon>0\) there is an \(N\) such that \(||a_n-a|-0| = |a_n-a|<\epsilon\) whenever \(n>N\) which is exactly what it means for \(a_n\to a\). We can simply reverse this argument to obtain the other direction.
Misc 2¶
Assume that \((a_n)\) converges to some \(a\in \mathbb{R}\) and \((b_n)\) converges to some \(b\in \mathbb{R}\). Prove that \((a_n\cdot b_n)\to ab\).
Answer
Although already proved in the book (Theorem 3.21), I would like to provide an alternative proof utilizing ideas from Misc 1, Exercise 3.18 and Exercise 3.8. To prove \((a_n\cdot b_n)\to ab\) is equivalent to proving \(|a_nb_n - ab|\to 0\) by Misc 1. Notice that
Thus, by the Squeeze Theorem it suffices to show that the expression \(|a_n-a|\cdot |b_n| + |a|\cdot |b_n-b|\) converges to zero. By Exercise 3.8, this is equivalent to showing that \(|a_n-a|\cdot |b_n|\to 0\) and \(|a|\cdot |b_n-b|\to 0\) individually.
-
Since \((a_n)\to a\) we have \(|a_n-a|\to 0\). Furthermore, \(|b_n|\) is bounded (since \(b_n\) is convergent), thus \(|a_n-a|\cdot |b_n|\to 0\) by Exercise 3.18.
-
Since \((b_n)\to b\) we have \(|b_n-b|\to 0\) so that \(|a||b_n-b|\to 0\) by Exercise 3.8.