Skip to content

§8 A Discussion about Proofs

This section works through several proofs that use the definition of the limit directly. With some study and practice you should be able to write proofs like these yourself. Each example has a discussion, which is the scratch work you would do on paper, followed by the formal proof you would hand in.

Intuitionthe shape of every ε–N proof

Every proof that sn→ss_n \to s has the same skeleton:

Let ε>0\varepsilon > 0. Let N=‾N = \underline{\quad\quad}. Then n>Nn > N implies ⋯\cdots, so ∣sn−s∣<ε|s_n - s| < \varepsilon.

The scratch work is a search for what to put in the blank, and it usually runs backward from the goal. The formal proof runs forward and only has to check that the chosen NN works. A reader should never have to see how you found it.

Example 1.

Prove lim⁡1n2=0\lim \dfrac{1}{n^2} = 0.

Discussion.

We must take an arbitrary ε>0\varepsilon > 0 and produce an NN, depending on ε\varepsilon, such that n>Nn > N implies ∣1n2−0∣<ε\left|\frac1{n^2} - 0\right| < \varepsilon. So the formal proof should begin “Let ε>0\varepsilon > 0” and end with “Hence n>Nn > N implies ∣1n2−0∣<ε\left|\frac1{n^2} - 0\right| < \varepsilon.” In between, it must name NN and check that it works.

To find NN, we work backward from what we want, as you might with a trigonometric identity, making sure every step is reversible. We want 1n2<ε\frac1{n^2} < \varepsilon. Multiplying by n2n^2 and dividing by ε\varepsilon, this is 1ε<n2\frac1\varepsilon < n^2, or 1ε<n\frac1{\sqrt\varepsilon} < n. This suggests N=1εN = \frac1{\sqrt\varepsilon}.

Formal Proof

Let ε>0\varepsilon > 0 and let N=1εN = \frac1{\sqrt\varepsilon}. Then n>Nn > N implies n>1εn > \frac1{\sqrt\varepsilon}, which implies n2>1εn^2 > \frac1\varepsilon and hence ε>1n2\varepsilon > \frac1{n^2}. Thus n>Nn > N implies ∣1n2−0∣<ε\left|\frac1{n^2} - 0\right| < \varepsilon. This proves lim⁡1n2=0\lim \frac1{n^2} = 0.

∎
Example 2.

Prove lim⁡3n+17n−4=37\lim \dfrac{3n+1}{7n-4} = \dfrac37.

Discussion.

For each ε>0\varepsilon > 0 we must decide how big nn must be to guarantee ∣3n+17n−4−37∣<ε\left|\frac{3n+1}{7n-4} - \frac37\right| < \varepsilon. Combining the fractions, we want

∣21n+7−21n+127(7n−4)∣<εor197(7n−4)<ε.\left| \frac{21n + 7 - 21n + 12}{7(7n - 4)} \right| < \varepsilon \qquad\text{or}\qquad \frac{19}{7(7n-4)} < \varepsilon.

Since 7n−4>07n - 4 > 0 we dropped the absolute value. Now solve for nn:

197ε<7n−4or197ε+4<7nor1949ε+47<n.\frac{19}{7\varepsilon} < 7n - 4 \qquad\text{or}\qquad \frac{19}{7\varepsilon} + 4 < 7n \qquad\text{or}\qquad \frac{19}{49\varepsilon} + \frac47 < n.

These steps are reversible, so we put N=1949ε+47N = \frac{19}{49\varepsilon} + \frac47. Any larger NN would also work.

Formal Proof

Let ε>0\varepsilon > 0 and let N=1949ε+47N = \frac{19}{49\varepsilon} + \frac47. Then n>Nn > N implies n>1949ε+47n > \frac{19}{49\varepsilon} + \frac47, hence 7n>197ε+47n > \frac{19}{7\varepsilon} + 4, hence 7n−4>197ε7n - 4 > \frac{19}{7\varepsilon}, hence 197(7n−4)<ε\frac{19}{7(7n-4)} < \varepsilon, and hence ∣3n+17n−4−37∣<ε\left|\frac{3n+1}{7n-4} - \frac37\right| < \varepsilon. This proves the claim.

∎

This is the NN that the ε–N widget in §7 uses for this sequence.

Example 3.

Prove lim⁡4n3+3nn3−6=4\lim \dfrac{4n^3 + 3n}{n^3 - 6} = 4.

Discussion.

We must find how large nn has to be to make

∣4n3+3nn3−6−4∣<εor∣3n+24n3−6∣<ε.\left| \frac{4n^3 + 3n}{n^3 - 6} - 4 \right| < \varepsilon \qquad\text{or}\qquad \left| \frac{3n + 24}{n^3 - 6} \right| < \varepsilon.

For n>1n > 1 we can drop the absolute values. This time it would be very hard to solve for nn. But we do not need the least NN, only some NN, so we simplify with estimates. The idea is that 3n+24n3−6\frac{3n+24}{n^3-6} is at most a constant times nn3=1n2\frac{n}{n^3} = \frac1{n^2} for large nn. To get such a bound we make the numerator bigger and the denominator smaller.

Since 3n+24≤27n3n + 24 \le 27n, it suffices to get 27nn3−6<ε\frac{27n}{n^3 - 6} < \varepsilon. Next, n3−6≥n32n^3 - 6 \ge \frac{n^3}{2} once n32≥6\frac{n^3}2 \ge 6, which holds for n>2n > 2. So for n>2n > 2 it suffices to get 27nn3/2<ε\frac{27n}{n^3/2} < \varepsilon, i.e. 54n2<ε\frac{54}{n^2} < \varepsilon, i.e. n>54/εn > \sqrt{54/\varepsilon}.

Formal Proof

Let ε>0\varepsilon > 0 and let N=max⁡{2,54/ε}N = \max\left\{2, \sqrt{54/\varepsilon}\right\}. Then n>Nn > N implies n>54/εn > \sqrt{54/\varepsilon}, hence 54n2<ε\frac{54}{n^2} < \varepsilon, hence 27nn3/2<ε\frac{27n}{n^3/2} < \varepsilon. Since n>2n > 2, we also have n32≤n3−6\frac{n^3}2 \le n^3 - 6, and 27n≥3n+2427n \ge 3n + 24. So n>Nn > N implies

3n+24n3−6≤27n12n3=54n2<ε,hence∣4n3+3nn3−6−4∣<ε.\frac{3n + 24}{n^3 - 6} \le \frac{27n}{\frac12 n^3} = \frac{54}{n^2} < \varepsilon, \qquad\text{hence}\qquad \left| \frac{4n^3 + 3n}{n^3 - 6} - 4 \right| < \varepsilon.∎

Intuitionhow to estimate

You want An<εA_n < \varepsilon, but AnA_n is too messy to solve for nn. Find a simpler BnB_n with An≤BnA_n \le B_n for all large nn, and make Bn<εB_n < \varepsilon instead. To make a fraction bigger, increase the numerator or decrease the denominator, never the reverse. Each simplification usually needs nn to be large enough, like n>2n > 2 above. Collect those requirements and build them into NN with a max⁡\max.

Common pitfall

  • When shrinking a denominator, keep it positive. n3−6≥n32n^3 - 6 \ge \frac{n^3}{2} is useful only because both sides are positive once n>2n > 2.
  • Backward scratch work is safe only if each step can be reversed. If one cannot, check the forward direction separately in the formal proof.
  • Every assumption about nn has to be justified by n>Nn > N. If your estimate needs n≥10n \ge 10, the formal proof needs N≥10N \ge 10.

As Example 3 shows, direct proofs of even simple limits can get messy. With the limit theorems of §9 we will just write

lim⁡4n3+3nn3−6=lim⁡4+3n21−6n3=4+3lim⁡1n21−6lim⁡1n3=4.\lim \frac{4n^3 + 3n}{n^3 - 6} = \lim \frac{4 + \frac3{n^2}}{1 - \frac6{n^3}} = \frac{4 + 3\lim\frac1{n^2}}{1 - 6\lim\frac1{n^3}} = 4.
Example 4.

Show that the sequence an=(−1)na_n = (-1)^n does not converge.

Discussion.

Assume lim⁡(−1)n=a\lim (-1)^n = a and look for a contradiction. Whatever aa is, at least one of 11 and −1-1 is at distance at least 11 from aa. So ∣(−1)n−a∣<1|(-1)^n - a| < 1 cannot hold for all large nn.

Formal Proof

Assume lim⁡(−1)n=a\lim(-1)^n = a for some a∈Ra \in \mathbb{R}. Taking ε=1\varepsilon = 1 in the definition of limit, there is NN such that n>Nn > N implies ∣(−1)n−a∣<1|(-1)^n - a| < 1. Using an even and an odd n>Nn > N, we get ∣1−a∣<1|1 - a| < 1 and ∣−1−a∣<1|-1 - a| < 1. By the Triangle Inequality 3.7,

2=∣1−(−1)∣=∣1−a+a−(−1)∣≤∣1−a∣+∣a−(−1)∣<1+1=2.2 = |1 - (-1)| = |1 - a + a - (-1)| \le |1 - a| + |a - (-1)| < 1 + 1 = 2.

This absurdity shows the assumption was wrong, so (−1)n(-1)^n does not converge.

∎

Intuitionproving divergence

To show sn↛as_n \not\to a, negate Definition 7.1: find one ε>0\varepsilon > 0 such that, however large NN is, some n>Nn > N has ∣sn−a∣≥ε|s_n - a| \ge \varepsilon. Example 4 does this with ε=1\varepsilon = 1 for every candidate aa at once. A single bad tolerance is enough.

Example 5.

Let (sn)(s_n) be a sequence of nonnegative real numbers with s=lim⁡sns = \lim s_n. Note s≥0s \ge 0 (Exercise 8.9(a)). Prove lim⁡sn=s\lim \sqrt{s_n} = \sqrt{s}.

Discussion.

This time we cannot expect a formula for NN in terms of ε\varepsilon, because the problem is so general. But we can show some NN exists. The trick is to break with algebra training and “irrationalize the denominator”:

sn−s=(sn−s)(sn+s)sn+s=sn−ssn+s.\sqrt{s_n} - \sqrt{s} = \frac{(\sqrt{s_n} - \sqrt s)(\sqrt{s_n} + \sqrt s)}{\sqrt{s_n} + \sqrt s} = \frac{s_n - s}{\sqrt{s_n} + \sqrt s}.

Since sn→ss_n \to s we can make the numerator small. If s>0s > 0 the denominator is at least s\sqrt s, so ∣sn−s∣≤∣sn−s∣s|\sqrt{s_n} - \sqrt s| \le \frac{|s_n - s|}{\sqrt s}, and we choose NN so that ∣sn−s∣<s ε|s_n - s| < \sqrt s\,\varepsilon for n>Nn > N. Such an NN exists because the definition of limit can be applied to s ε\sqrt s\,\varepsilon just as well as to ε\varepsilon. If s=0s = 0 the denominator can be small too, but then a direct argument works (Exercise 8.3).

Formal Proof

Case I: s>0s > 0. Let ε>0\varepsilon > 0. Since lim⁡sn=s\lim s_n = s, there is NN such that n>Nn > N implies ∣sn−s∣<s ε|s_n - s| < \sqrt s\,\varepsilon. Then n>Nn > N implies

∣sn−s∣=∣sn−s∣sn+s≤∣sn−s∣s<s εs=ε.|\sqrt{s_n} - \sqrt s| = \frac{|s_n - s|}{\sqrt{s_n} + \sqrt s} \le \frac{|s_n - s|}{\sqrt s} < \frac{\sqrt s\,\varepsilon}{\sqrt s} = \varepsilon.

Case II: s=0s = 0. This is Exercise 8.3.

∎
Example 6.

Let (sn)(s_n) be a convergent sequence with sn≠0s_n \ne 0 for all nn and lim⁡sn=s≠0\lim s_n = s \ne 0. Prove inf⁡{∣sn∣:n∈N}>0\inf\{|s_n| : n \in \mathbb{N}\} > 0.

Discussion.

“Most” terms sns_n are close to ss and hence not close to 00. More precisely, most sns_n are within 12∣s∣\frac12|s| of ss, so most satisfy ∣sn∣≥12∣s∣|s_n| \ge \frac12 |s|. The finitely many remaining terms are nonzero, so they cannot drag the infimum down to 00.

Formal Proof

Let ε=12∣s∣>0\varepsilon = \frac12|s| > 0. Since lim⁡sn=s\lim s_n = s, there is N∈NN \in \mathbb{N} such that n>Nn > N implies ∣sn−s∣<∣s∣2|s_n - s| < \frac{|s|}2. Then

n>Nimplies∣sn∣≥∣s∣2,(1)n > N \quad\text{implies}\quad |s_n| \ge \frac{|s|}{2}, \tag{1}

since otherwise the Triangle Inequality would give ∣s∣=∣s−sn+sn∣≤∣s−sn∣+∣sn∣<∣s∣2+∣s∣2=∣s∣|s| = |s - s_n + s_n| \le |s - s_n| + |s_n| < \frac{|s|}2 + \frac{|s|}2 = |s|, which is absurd. Put m=min⁡{∣s∣2,∣s1∣,∣s2∣,…,∣sN∣}m = \min\left\{\frac{|s|}2, |s_1|, |s_2|, \dots, |s_N|\right\}. Then m>0m > 0 and, by (1), ∣sn∣≥m|s_n| \ge m for all nn. So inf⁡{∣sn∣:n∈N}≥m>0\inf\{|s_n| : n \in \mathbb{N}\} \ge m > 0.

∎

Reading alongside Abbott

Abbott’s Section 2.2 works through similar examples and stresses the same two stages: rough work to find NN, then a clean forward proof that starts with “Let ε>0\varepsilon > 0”.

This is the classic slip in turning scratch work into a proof.

Spot the error: the limit of 1/n

One line of this argument is wrong. Press the button on the line you think is the mistake.

  1. Claim: lim⁡1n=0\lim \frac1n = 0. Let ε>0\varepsilon > 0.
  2. Suppose that ∣1n−0∣<ε\left|\frac1n - 0\right| < \varepsilon.
  3. Then 1n<ε\frac1n < \varepsilon, which means n>1εn > \frac1\varepsilon.
  4. So take N=1εN = \frac1\varepsilon. Hence lim⁡1n=0\lim \frac1n = 0.

Take the quiz (3 questions)

To prove lim⁡1n2=0\lim \frac{1}{n^2} = 0, given ε>0\varepsilon > 0, which choice of NN works?

Your scratch work shows ∣sn−s∣<ε|s_n - s| < \varepsilon whenever n>3εn > \frac3\varepsilon. How should the formal proof begin?

If lim⁡sn=s\lim s_n = s and s>0s > 0, which follows?

Formal proofs are required in the following exercises.

Exercise 8.1

Prove the following:

  1. lim⁡(−1)nn=0\lim \dfrac{(-1)^n}{n} = 0
  2. lim⁡1n1/3=0\lim \dfrac{1}{n^{1/3}} = 0
  3. lim⁡2n−13n+2=23\lim \dfrac{2n-1}{3n+2} = \dfrac23
  4. lim⁡n+6n2−6=0\lim \dfrac{n+6}{n^2-6} = 0
Hint

(a) N=1εN = \frac1\varepsilon. (b) N=1ε3N = \frac1{\varepsilon^3}. (c) ∣2n−13n+2−23∣=73(3n+2)<79n\left|\frac{2n-1}{3n+2} - \frac23\right| = \frac{7}{3(3n+2)} < \frac{7}{9n}, so N=79εN = \frac{7}{9\varepsilon} works. (d) For n≥6n \ge 6: n+6≤2nn + 6 \le 2n and n2−6≥n22n^2 - 6 \ge \frac{n^2}2, so the term is at most 4n\frac4n. Take N=max⁡{6,4ε}N = \max\{6, \frac4\varepsilon\}.

Solution

In each part let ε>0\varepsilon > 0 and find NN so that n>Nn > N implies ∣sn−s∣<ε|s_n - s| < \varepsilon.

  1. ∣(−1)nn−0∣=1n<ε\left|\frac{(-1)^n}{n} - 0\right| = \frac1n < \varepsilon whenever n>N=1εn > N = \frac1\varepsilon.
  2. 1n1/3<ε\frac1{n^{1/3}} < \varepsilon iff n1/3>1εn^{1/3} > \frac1\varepsilon iff n>1ε3n > \frac1{\varepsilon^3}; take N=1ε3N = \frac1{\varepsilon^3}.
  3. 2n−13n+2−23=3(2n−1)−2(3n+2)3(3n+2)=−73(3n+2)\dfrac{2n-1}{3n+2} - \dfrac23 = \dfrac{3(2n-1) - 2(3n+2)}{3(3n+2)} = \dfrac{-7}{3(3n+2)}, so the distance is 73(3n+2)<79n\dfrac{7}{3(3n+2)} < \dfrac7{9n}. This is <ε< \varepsilon when n>N=79εn > N = \frac7{9\varepsilon}.
  4. For n≥6n \ge 6 we have n+6≤2nn + 6 \le 2n and n2−6≥n22n^2 - 6 \ge \frac{n^2}2 (since n2≥12n^2 \ge 12), so 0<n+6n2−6≤2nn2/2=4n0 < \frac{n+6}{n^2-6} \le \frac{2n}{n^2/2} = \frac4n. This is <ε< \varepsilon for n>N=max⁡{6,4ε}n > N = \max\{6, \frac4\varepsilon\}.

Tags: Standardε–N / ε–δ

Exercise 8.2

Determine the limits of the following sequences, and then prove your claims.

  1. an=nn2+1a_n = \dfrac{n}{n^2+1}
  2. bn=7n−193n+7b_n = \dfrac{7n-19}{3n+7}
  3. cn=4n+37n−5c_n = \dfrac{4n+3}{7n-5}
  4. dn=2n+45n+2d_n = \dfrac{2n+4}{5n+2}
  5. sn=1nsin⁡ns_n = \dfrac1n \sin n
Solution
  1. 00
  2. 73\tfrac73
  3. 47\tfrac47
  4. 25\tfrac25
  5. 00

For (b), for example: ∣7n−193n+7−73∣=1063(3n+7)<1069n\left|\frac{7n-19}{3n+7} - \frac73\right| = \frac{106}{3(3n+7)} < \frac{106}{9n}, so take N=1069εN = \frac{106}{9\varepsilon}. For (e), ∣sin⁡nn∣≤1n\left|\frac{\sin n}{n}\right| \le \frac1n.

Tags: Standardε–N / ε–δ

Exercise 8.3

Let (sn)(s_n) be a sequence of nonnegative real numbers with lim⁡sn=0\lim s_n = 0. Prove lim⁡sn=0\lim \sqrt{s_n} = 0. This completes the proof of Example 5.

Hint

Given ε>0\varepsilon > 0, apply the definition of lim⁡sn=0\lim s_n = 0 with ε2\varepsilon^2 in place of ε\varepsilon.

Solution

Let ε>0\varepsilon > 0. Since sn→0s_n \to 0, applying the definition with ε2\varepsilon^2 gives NN such that n>Nn > N implies sn<ε2s_n < \varepsilon^2 (recall sn≥0s_n \ge 0). Then 0≤sn<ε0 \le \sqrt{s_n} < \varepsilon for n>Nn > N, so sn→0\sqrt{s_n} \to 0.

Tags: Standardε–N / ε–δ

Exercise 8.4

Let (tn)(t_n) be a bounded sequence, i.e. there is MM with ∣tn∣≤M|t_n| \le M for all nn, and let (sn)(s_n) be a sequence with lim⁡sn=0\lim s_n = 0. Prove lim⁡(sntn)=0\lim (s_n t_n) = 0.

Hint

We may take M>0M > 0. Choose NN so that ∣sn∣<εM|s_n| < \frac{\varepsilon}{M} for n>Nn > N.

Solution

We may assume M>0M > 0. Let ε>0\varepsilon > 0 and choose NN so that ∣sn∣<εM|s_n| < \frac\varepsilon M for n>Nn > N. Then for n>Nn > N, ∣sntn−0∣=∣sn∣∣tn∣≤M∣sn∣<ε.|s_nt_n - 0| = |s_n||t_n| \le M|s_n| < \varepsilon .

Tags: Standardε–N / ε–δ

Exercise 8.5

This exercise is referred to in several places.

  1. Consider sequences (an)(a_n), (bn)(b_n), (sn)(s_n) with an≤sn≤bna_n \le s_n \le b_n for all nn and lim⁡an=lim⁡bn=s\lim a_n = \lim b_n = s. Prove lim⁡sn=s\lim s_n = s. This is the squeeze lemma.
  2. Suppose ∣sn∣≤tn|s_n| \le t_n for all nn and lim⁡tn=0\lim t_n = 0. Prove lim⁡sn=0\lim s_n = 0.
Hint

(a) For nn beyond both NN’s, s−ε<an≤sn≤bn<s+εs - \varepsilon < a_n \le s_n \le b_n < s + \varepsilon. (b) Apply (a) with an=−tna_n = -t_n and bn=tnb_n = t_n.

Solution
  1. Let ε>0\varepsilon > 0. Choose N1N_1 with ∣an−s∣<ε|a_n - s| < \varepsilon for n>N1n > N_1 and N2N_2 with ∣bn−s∣<ε|b_n - s| < \varepsilon for n>N2n > N_2. For n>max⁡{N1,N2}n > \max\{N_1, N_2\}, s−ε<an≤sn≤bn<s+ε,s - \varepsilon < a_n \le s_n \le b_n < s + \varepsilon, so ∣sn−s∣<ε|s_n - s| < \varepsilon.
  2. From ∣sn∣≤tn|s_n| \le t_n we get −tn≤sn≤tn-t_n \le s_n \le t_n. Since −tn→0-t_n \to 0 and tn→0t_n \to 0, part (a) gives sn→0s_n \to 0.

Tags: Standardε–N / ε–δ

Used in Theorem 9.7, Theorem 11.2

Exercise 8.6

Let (sn)(s_n) be a sequence in R\mathbb{R}.

  1. Prove lim⁡sn=0\lim s_n = 0 if and only if lim⁡∣sn∣=0\lim |s_n| = 0.
  2. Observe that if sn=(−1)ns_n = (-1)^n, then lim⁡∣sn∣\lim |s_n| exists, but lim⁡sn\lim s_n does not.
Hint

(a) ∣sn−0∣=∣∣sn∣−0∣|s_n - 0| = \bigl||s_n| - 0\bigr|, so the two definitions say exactly the same thing.

Solution
  1. ∣sn−0∣=∣sn∣=∣∣sn∣−0∣|s_n - 0| = |s_n| = \bigl||s_n| - 0\bigr| for every nn, so “∣sn−0∣<ε|s_n - 0| < \varepsilon for n>Nn > N” and “∣∣sn∣−0∣<ε\bigl||s_n| - 0\bigr| < \varepsilon for n>Nn > N” are the same statement. Hence sn→0s_n \to 0 iff ∣sn∣→0|s_n| \to 0.
  2. If sn=(−1)ns_n = (-1)^n then ∣sn∣=1|s_n| = 1 for all nn, so lim⁡∣sn∣=1\lim|s_n| = 1 exists, while lim⁡(−1)n\lim(-1)^n does not exist (Example 4). (Part (a) is special to the limit 00.)

Tags: Standardε–N / ε–δ

Used in Exercise 12.2

Exercise 8.7

Show the following sequences do not converge.

  1. cos⁡(nπ3)\cos\left(\dfrac{n\pi}{3}\right)
  2. sn=(−1)nns_n = (-1)^n n
  3. sin⁡(nπ3)\sin\left(\dfrac{n\pi}{3}\right)
Hint

Imitate Example 4. (a) The values 11 (at n=6kn = 6k) and −1-1 (at n=6k+3n = 6k + 3) recur forever, so take ε=1\varepsilon = 1. (b) A convergent sequence is bounded (Theorem 9.1), or argue directly. (c) 32\frac{\sqrt3}{2} and −32-\frac{\sqrt3}{2} recur forever; take ε=32\varepsilon = \frac{\sqrt3}{2}.

Solution
  1. For n=6kn = 6k the term is cos⁡(2kπ)=1\cos(2k\pi) = 1 and for n=6k+3n = 6k+3 it is cos⁡((2k+1)π)=−1\cos((2k+1)\pi) = -1; both values occur for arbitrarily large nn. Suppose the limit were aa. Taking ε=1\varepsilon = 1, for all large nn we’d have ∣sn−a∣<1|s_n - a| < 1, and choosing large nn with sn=1s_n = 1 and large mm with sm=−1s_m = -1 gives 2=∣1−(−1)∣≤∣1−a∣+∣a+1∣<22 = |1 - (-1)| \le |1 - a| + |a + 1| < 2, a contradiction.
  2. ∣sn∣=n|s_n| = n is unbounded, but convergent sequences are bounded (Theorem 9.1). So (sn)(s_n) does not converge.
  3. For n=6k+1n = 6k+1 the term is sin⁡(π3+2kπ)=32\sin(\frac\pi3 + 2k\pi) = \frac{\sqrt3}2, and for n=6k+4n = 6k+4 it is sin⁡(4π3+2kπ)=−32\sin(\frac{4\pi}3 + 2k\pi) = -\frac{\sqrt3}2; both recur forever. If sn→as_n \to a, take ε=32\varepsilon = \frac{\sqrt3}2: large nn and mm with these values give 3≤∣sn−a∣+∣a−sm∣<3\sqrt3 \le |s_n - a| + |a - s_m| < \sqrt3, a contradiction.

Tags: Standardε–N / ε–δContradiction

Used in §7

Exercise 8.8

Prove the following (see Exercise 7.5):

  1. lim⁡[n2+1−n]=0\lim\left[\sqrt{n^2+1} - n\right] = 0
  2. lim⁡[n2+n−n]=12\lim\left[\sqrt{n^2+n} - n\right] = \frac12
  3. lim⁡[4n2+n−2n]=14\lim\left[\sqrt{4n^2+n} - 2n\right] = \frac14
Hint

(a) n2+1−n=1n2+1+n<12n\sqrt{n^2+1} - n = \frac{1}{\sqrt{n^2+1}+n} < \frac1{2n}. (b) ∣nn2+n+n−12∣=n2+n−n2(n2+n+n)<18n\left|\frac{n}{\sqrt{n^2+n}+n} - \frac12\right| = \frac{\sqrt{n^2+n} - n}{2(\sqrt{n^2+n}+n)} < \frac{1}{8n}, because n2+n−n=nn2+n+n<12\sqrt{n^2+n} - n = \frac{n}{\sqrt{n^2+n}+n} < \frac12 and the denominator exceeds 4n4n. (c) is similar.

Solution
  1. 0<n2+1−n=1n2+1+n<12n0 < \sqrt{n^2+1} - n = \dfrac{1}{\sqrt{n^2+1} + n} < \dfrac1{2n}, which is <ε< \varepsilon for n>12εn > \frac1{2\varepsilon}.
  2. Write n2+n−n=nn2+n+n\sqrt{n^2+n} - n = \dfrac{n}{\sqrt{n^2+n} + n}. Since n2+n<(n+12)2n^2 + n < (n + \frac12)^2, we have n2+n−n<12\sqrt{n^2+n} - n < \frac12. Then ∣n2+n−n−12∣=∣nn2+n+n−12∣=n2+n−n2(n2+n+n)<1/24n=18n,\left|\sqrt{n^2+n} - n - \tfrac12\right| = \left|\frac{n}{\sqrt{n^2+n}+n} - \frac12\right| = \frac{\sqrt{n^2+n} - n}{2(\sqrt{n^2+n} + n)} < \frac{1/2}{4n} = \frac1{8n}, because n2+n+n>2n\sqrt{n^2+n} + n > 2n. This is <ε< \varepsilon for n>18εn > \frac1{8\varepsilon}.
  3. Let T=4n2+nT = \sqrt{4n^2+n}. Then T−2n=nT+2nT - 2n = \dfrac{n}{T + 2n} and T>2nT > 2n, so T−2n<n4n=14T - 2n < \frac n{4n} = \frac14. Also ∣T−2n−14∣=∣nT+2n−14∣=T−2n4(T+2n)<1/416n=164n,\left|T - 2n - \tfrac14\right| = \left|\frac{n}{T+2n} - \frac14\right| = \frac{T - 2n}{4(T + 2n)} < \frac{1/4}{16n} = \frac1{64n}, which is <ε< \varepsilon for n>164εn > \frac1{64\varepsilon}.

Tags: Standardε–N / ε–δComputation

Exercise 8.9

Let (sn)(s_n) be a sequence that converges. (This exercise is referred to in several places.)

  1. Show that if sn≥as_n \ge a for all but finitely many nn, then lim⁡sn≥a\lim s_n \ge a.
  2. Show that if sn≤bs_n \le b for all but finitely many nn, then lim⁡sn≤b\lim s_n \le b.
  3. Conclude that if all but finitely many sns_n belong to [a,b][a, b], then lim⁡sn\lim s_n belongs to [a,b][a, b].
Hint

(a) Suppose s=lim⁡sn<as = \lim s_n < a. Take ε=a−s\varepsilon = a - s. Then for large nn, sn<s+ε=as_n < s + \varepsilon = a, which contradicts the hypothesis for all large nn.

Solution

Let s=lim⁡sns = \lim s_n.

  1. Suppose s<as < a and let ε=a−s>0\varepsilon = a - s > 0. For all large nn, sn<s+ε=as_n < s + \varepsilon = a, so sn<as_n < a for infinitely many nn (all large ones), contradicting that sn≥as_n \ge a for all but finitely many nn. Hence s≥as \ge a.
  2. Suppose s>bs > b and let ε=s−b\varepsilon = s - b. For all large nn, sn>s−ε=bs_n > s - \varepsilon = b, contradicting sn≤bs_n \le b for all but finitely many nn. Hence s≤bs \le b.
  3. If all but finitely many sns_n lie in [a,b][a,b], then sn≥as_n \ge a and sn≤bs_n \le b for all but finitely many nn, so a≤s≤ba \le s \le b by (a) and (b).

Tags: Standardε–N / ε–δContradiction

Used in §9, Exercise 9.4, Exercise 9.9, §10, Exercise 11.9, Theorem 18.1

Exercise 8.10

Let (sn)(s_n) be a convergent sequence with lim⁡sn>a\lim s_n > a. Prove there is a number NN such that n>Nn > N implies sn>as_n > a.

Hint

Use ε=lim⁡sn−a\varepsilon = \lim s_n - a.

Solution

Let s=lim⁡sn>as = \lim s_n > a and ε=s−a>0\varepsilon = s - a > 0. There is NN such that n>Nn > N implies ∣sn−s∣<ε|s_n - s| < \varepsilon, so sn>s−ε=as_n > s - \varepsilon = a.

Tags: Standardε–N / ε–δ

Used in Theorem 9.9, Exercise 9.11

Keyboard shortcuts

Single-key shortcuts
KeyAction
nNext section
pPrevious section
jNext exercise
kPrevious exercise
hShow or hide all hints
sShow or hide all solutions
/Search
?Show this list
Ctrl K or ⌘ KSearch (built in)
EscClose a dialog

Shortcuts never fire while you type in a text field or hold Ctrl, Alt or ⌘. Screen reader users in browse mode can press the keys after switching to focus mode, or turn them off here.