Study Guides / Calculus I / Exercise session 2
EXERCISE SESSION 2 · ETH ZURICH · FS20 · calculus

Convergence, Limits and Monotonicity

Convergence, divergence, limit, dominance, limit rules, monotonicity, boundedness of a sequence

SB
Steven Battilana Quant · ex-ETH

Convergence and Divergence of a Sequence

Definition 1.1 limit of a sequence I

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} converges with limit aa (converges to a∈Ra \in \mathbb{R}) if for every ε>0\varepsilon > 0 there is an index N(ε)≥1N(\varepsilon) \geq 1 such that

∣an−a∣<ε∀n>N(ε).|a_n -a| < \varepsilon \quad \forall n > N(\varepsilon).

Notation: lim⁡n→∞an=a\quad \lim_{n \to \infty} a_n = a.

Definition 1.2 limit of a sequence II

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} converges to a∈Ra \in \mathbb{R} if for every ε>0\varepsilon>0 the set of indices n≥1n\geq1 for which an∉(a−ε,a+ε)a_n \notin (a - \varepsilon, a + \varepsilon) is finite.

Epsilon neighbourhood around the limit a: only finitely many terms of the sequence lie outside
Definition 1.3

A sequence is called convergent if it has a limit; otherwise it is called divergent.

Example 1.1
  1. lim⁡n→±∞n\lim\limits _{n \to \pm \infty} n
    Show solutionHide solution lim⁡n→±∞n=±∞\lim\limits _{n \to \pm \infty} n = \pm \infty
  2. lim⁡n→∞(−1)n\lim\limits _{n \to \infty} (-1)^n
    Show solutionHide solution lim⁡n→∞(−1)n=\lim\limits _{n \to \infty} (-1)^n = diverges
Example 1.2

To show: prove with the definition that the following holds: lim⁡n→∞1n=0\quad \lim_{n \to \infty} \frac{1}{n} = 0

Show solutionHide solution

Proof:

We choose an arbitrarily small ε>0\varepsilon > 0 (e.g. ε=10−5\varepsilon = 10^{-5} or ε=10−10\varepsilon = 10^{-10}). By the definition, we have to show that there exists an N=N(ε)∈NN = N(\varepsilon) \in \mathbb{N} such that for all n>Nn > N the following holds:

∣an−a∣=∣1n−0∣=∣1n∣<ε.|a_n-a| = \left| \frac{1}{n}-0 \right| = \left| \frac{1}{n} \right| < \varepsilon.

How do you do that? We solve for nn

∣an−a∣=∣1n∣<!ε⇒n>0n>1ε.|a_n-a| = \left| \frac{1}{n} \right| \stackrel{!}{<} \varepsilon \quad \stackrel{n>0}{\Rightarrow} \quad n > \frac{1}{\varepsilon}.

We therefore choose N:=⌈1ε⌉N:= \left\lceil \frac{1}{\varepsilon} \right\rceil, where ⌈⋅⌉\lceil \cdot \rceil is the ceiling function (e.g. ⌈123.12⌉\lceil 123.12 \rceil = 124). Hence for n>Nn > N we have (by construction):

∣1n−0∣<ε.\left| \frac{1}{n}-0 \right| < \varepsilon.

This is exactly the definition of lim⁡n→∞1n=0\quad \lim_{n \to \infty} \frac{1}{n} = 0.

Satz

If a limit exists, it is unique.

Computing with Limits

Remark dominance

The following dominance relations hold:

  1. For x→∞:1≤log(x)≤x \to \infty: \quad 1 \leq log(x) \leq x≤xn  (for n≥12)\sqrt{x} \leq x^n \; (\text{for } n \geq \tfrac{1}{2}) ≤nx  (for n>1)≤x!≤xx\leq n^x \; (\text{for } n>1) \leq x! \leq x^x
  2. For x→0:log(x)≤xn≤1xnx \to 0: \quad log(x) \leq x^n \leq \frac{1}{x^n}
Satz rules for limits

If (an)n∈N(a_n)_{n \in \mathbb{N}} and (bn)n∈N(b_n)_{n \in \mathbb{N}} are convergent with limits aa and bb respectively, then:

  1. lim⁡n→∞(an+bn)=a+b\lim\limits_{n \to \infty} (a_n + b_n) = a+b
  2. lim⁡n→∞an⋅bn=a⋅b\lim\limits_{n \to \infty} a_n \cdot b_n = a \cdot b
  3. If bn,  b≠0b_n, \; b \neq 0, then: lim⁡n→∞anbn=ab\quad \lim\limits_{n \to \infty} \frac{a_n}{b_n} = \frac{a}{b}
  4. If an≤bn,  ∀n∈Na_n \leq b_n, \; \forall n \in \mathbb{N}, then: a≤b\quad a \leq b
  5. If an,  a>0a_n, \; a > 0, then: lim⁡n→∞anbn=ab\quad \lim\limits_{n \to \infty} a_n^{b_n} = a^b
  6. lim⁡n→∞f(an)=f(a),∀f\lim\limits_{n \to \infty} f(a_n) = f(a), \quad \forall f continuous.
Example 2.1

Investigate the convergence behaviour of

an=(1n+n2)9951+n1990.a_n = \frac{(\frac{1}{n}+n^2)^{995}}{1+n^{1990}}.
Show solutionHide solution

Since neither the denominator nor the numerator converges, the third rule (iii) cannot be applied directly. We have to rewrite the terms as follows to get around the problem.

(1n+n2)9951+n1990=(n2(1n3+1))9951+n1990=n1990(1n3+1)995n1990(1n1990+1)=(1n3+1)9951n1990+1\begin{aligned} \frac{(\frac{1}{n}+n^2)^{995}}{1+n^{1990}} &= \frac{(n^2(\frac{1}{n^3}+1))^{995}}{1+n^{1990}}\\ &= \frac{n^{1990}(\frac{1}{n^3}+1)^{995}}{n^{1990}(\frac{1}{n^{1990}}+1)}\\ &= \frac{(\frac{1}{n^3}+1)^{995}}{\frac{1}{n^{1990}}+1} \end{aligned}

Since 1n3→n→∞0\frac{1}{n^3} \xrightarrow{n \to \infty} 0 and 1n1990→n→∞0\frac{1}{n^{1990}} \xrightarrow{n \to \infty} 0, we obtain

lim⁡n→∞(1n3+1)9951n1990+1=(0+1)9950+1=19951=1.\lim_{n \to \infty} \frac{(\frac{1}{n^3}+1)^{995}}{\frac{1}{n^{1990}}+1} = \frac{(0+1)^{995}}{0+1} = \frac{1^{995}}{1} = 1.
Satz sandwich theorem

Let three sequences an≤bn≤cna_n \leq b_n \leq c_n be given. If ana_n and cnc_n converge with lim⁡n→∞an=lim⁡n→∞cn=L\lim\limits_{n \to \infty} a_n = \lim\limits_{n \to \infty} c_n = L with L∈RL \in \mathbb{R}, then bnb_n converges as well and

lim⁡n→∞bn=L.\lim_{n \to \infty} b_n = L.
Example 2.2

Compute lim⁡n→∞2n2n\lim\limits_{n \to \infty} \frac{2n}{2^n}.

Show solutionHide solution

For n≥1n \geq 1 we have 2n≥12n \geq 1, hence:

2n2n≥12n.\frac{2n}{2^n} \geq \frac{1}{2^n}.

For n≥4n \geq 4 we have 2n≥n22^n \geq n^2, hence:

2n2n≤2nn2=2n.\frac{2n}{2^n} \leq \frac{2n}{n^2} = \frac{2}{n}.

So for n≥4n \geq 4 we obtain the following estimates:

12n≤2n2n≤2n\frac{1}{2^n} \leq \frac{2n}{2^n} \leq \frac{2}{n}

The right-hand and left-hand sides converge to 00:

lim⁡n→∞12n=lim⁡n→∞2n=0⟹sandwich thmlim⁡n→∞2n2n=0.\begin{aligned} \lim_{n \to \infty} \frac{1}{2^n} = \lim_{n \to \infty} \frac{2}{n} &= 0\\ \stackrel{\text{sandwich thm}}{\Longrightarrow} \qquad \qquad \lim\limits_{n \to \infty} \frac{2n}{2^n} &= 0. \end{aligned}

Monotonicity and Convergence

Definition 3.1

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} is called monotonically increasing if for all n∈Nn \in \mathbb{N}:

an+1≥an.a_{n+1} \geq a_n.
Definition 3.2

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} is called strictly monotonically increasing if for all n∈Nn \in \mathbb{N}:

an+1>an.a_{n+1} > a_n.
Definition 3.3

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} is called monotonically decreasing if for all n∈Nn \in \mathbb{N}:

an+1≤an.a_{n+1} \leq a_n.
Definition 3.4

A sequence (an)n∈N(a_n)_{n \in \mathbb{N}} is called strictly monotonically decreasing if for all n∈Nn \in \mathbb{N}:

an+1<an.a_{n+1} < a_n.
Remark

To show monotonicity, you can proceed as follows: replace nn by the continuous variable xx and compute the derivative with respect to xx. If a′(x)≥0a'(x) \geq 0 or a′(x)≤0a'(x) \leq 0 respectively, the sequence is monotonically increasing or monotonically decreasing respectively.

Remark

Another option is to show directly that an+1an>1\frac{a_{n+1}}{a_n} >1 (only if an>0a_n > 0 for all nn) or an+1−an>0a_{n+1} - a_n > 0, analogously for the other cases.

Definition 3.5 (3.3.1.)

(an)n∈N(a_n)_{n \in \mathbb{N}} is called bounded above (below) if

∃b∈R,  ∀n∈N:an≤b(or b≤an);\exists b \in \mathbb{R}, \; \forall n \in \mathbb{N}: \quad a_n \leq b \qquad (\text{or } b \leq a_n);

that is, if the set A={an∣n∈N}A = \{a_n | n \in \mathbb{N}\} is bounded above (below).

Satz (3.3.1.)

If (an)n∈N(a_n)_{n \in \mathbb{N}} is convergent, then (an)n∈N(a_n)_{n \in \mathbb{N}} is bounded.

Remark

Boundedness is therefore necessary, but not sufficient, for convergence, as the example of the sequence an=(−1)n,  n∈Na_n = (-1)^n, \; n \in \mathbb{N}, shows.

an convergent⇒an boundedBUT:an convergent⇍an bounded\begin{aligned} a_n \text{ convergent} &\Rightarrow a_n \text{ bounded}\\ \text{BUT:} \quad a_n \text{ convergent} &\boldsymbol{\nLeftarrow} a_n \text{ bounded} \end{aligned}
Satz monotone convergence theorem (3.3.2.)

Let (an)n∈N(a_n)_{n \in \mathbb{N}} be bounded above and monotonically increasing, that is, for some number b∈Rb \in \mathbb{R} we have

∀n∈N:a1≤...≤an≤an+1≤...≤b.\forall n \in \mathbb{N}: \quad a_1 \leq ... \leq a_n \leq a_{n+1} \leq ... \leq b.

Then (an)n∈N(a_n)_{n \in \mathbb{N}} is convergent, and lim⁡n→∞an=sup⁡n∈Nan≤b\lim\limits_{n \to \infty} a_n = \sup\limits_{n \in \mathbb{N}} a_n \leq b.
Analogously if (an)n∈N(a_n)_{n \in \mathbb{N}} is bounded below and monotonically decreasing.

Remark

The rule of thumb is:
boundedness ++ monotonicity == convergence.

Example 3.1

Consider the recursively defined sequence

a0=0,an+1=(an2)2+1a_0 = 0, \quad a_{n+1} = \left( \frac{a_n}{2} \right)^2 +1

Show that the sequence ana_n converges. What is the limit?

Show solutionHide solution

We compute a few terms to get a feel for how the sequence behaves

a0=0,a1=1,a2=54=1.25,a3=8964≈1.39,...a_0 = 0, \quad a_1 = 1, \quad a_2 = \frac{5}{4} = 1.25, \quad a_3 = \frac{89}{64} \approx 1.39, \quad ...

The numbers suggest that the sequence is monotonically increasing.

Monotonicity:
Instead of showing an+1≥ana_{n+1} \geq a_n directly, we show the equivalent statement an+1−an≥0a_{n+1} - a_n \geq 0. We have

an+1−an=an24+1−an=an2−4an+44=(an−2)24≥0\begin{aligned} a_{n+1} - a_n &= \frac{a_n^2}{4}+1 - a_n\\ &= \frac{a_n^2 - 4a_n +4}{4}\\ &= \frac{(a_n-2)^2}{4} \geq 0 \end{aligned}

Boundedness:
We show by induction that an≤2  ∀n∈Na_n \leq 2 \; \forall n \in \mathbb{N}. (Assuming a limit exists, you can compute it with the last step, but then you still have to show boundedness.)

Base case (n=0)(n=0):

a0=0≤2✓a_0 = 0 \leq 2 \quad \checkmark

Induction step (n↦n+1)(n \mapsto n+1):
We assume that an≤2a_n \leq 2 holds for a fixed but arbitrary n∈Nn \in \mathbb{N} (induction hypothesis, IH). It follows that

an+1=an24+1≤IH,  0≤an224+1=2.a_{n+1} = \frac{a_n^2}{4} +1 \stackrel{\text{IH},\; 0 \leq a_n}{\leq} \frac{2^2}{4} +1 = 2.

(0≤an0 \leq a_n holds, since a0=0a_0 = 0 and an+1=an24+1≥1a_{n+1} = \frac{a_n^2}{4} + 1 \geq 1.)

Now we can apply the monotone convergence theorem. It follows that ana_n converges, i.e. lim⁡n→∞an=a\lim\limits_{n \to \infty} a_n = a. Now we compute the limit aa.

an+1=(an2)2+1⟹an+1→aan→aa=a24+1⇔(a−2)24=0⇔(a−2)2=0⇔a=2.\begin{aligned} a_{n+1} = \left( \frac{a_n}{2} \right)^2 +1 &\overset{\substack{a_{n+1} \rightarrow a\\ a_n \rightarrow a}}{\Longrightarrow} a = \frac{a^2}{4} +1\\ &\Leftrightarrow \frac{(a-2)^2}{4} = 0\\ &\Leftrightarrow (a-2)^2 = 0\\ &\Leftrightarrow a = 2. \end{aligned}