Week 8 Discussion Notes

Table of Contents

Homework Comments

9.3

Here is the problem statement:

Suppose limnan=a\displaystyle\lim_{n\to\infty} a_n = a, limnbn=b\displaystyle\lim_{n\to\infty} b_n = b, and sn=an3+4anbn2+1s_n = \dfrac{a_n^3+4a_n}{b_n^2+1}. Prove limnsn=a3+4ab2+a\displaystyle\lim_{n\to\infty} s_n = \frac{a^3 + 4a}{b^2 + a} carefully, using the limit theorems.

I emphasized the word "carefully" because it's the reason why I was fairly strict with grading. Many of you had something like this as the first line:

limnsn=limn(an3+4an)limn(bn2+1)(Theorem 9.6)\lim_{n\to\infty} s_n = \frac{\lim_{n\to\infty} \p{a_n^3 + 4a_n}}{\lim_{n\to\infty} \p{b_n^2 + 1}} \quad \p{\text{Theorem 9.6}}

Theorem 9.6 says the following:

If {sn}n\set{s_n}_n converges to ss and {tn}n\set{t_n}_n converges to tt with t0t \neq 0, then {tnsn}n\set{\frac{t_n}{s_n}}_n converges to ts\frac{t}{s}.

Remember that you need to check assumptions before using a fact. The assumptions for Theorem 9.6 are that limnsn\displaystyle\lim_{n\to\infty} s_n and limntn\displaystyle\lim_{n\to\infty} t_n exist. In other words, if you wrote the line above before checking that the limits of the numerator and denominator exist, you used Theorem 9.6 before checking all assumptions. So while you get the "correct" answer in the end, the reasoning is wrong, which is why I took off a lot of points.

To prove this properly, Theorem 9.6 turns out to be the last limit law that you use in this problem. A correct solution would look like the following:

By Theorem 9.4,

limnan3=(limnan)3=a3andlimnbn2=(limnbn)2=b2,\lim_{n\to\infty} a_n^3 = \p{\lim_{n\to\infty} a_n}^3 = a^3 \quad\text{and}\quad \lim_{n\to\infty} b_n^2 = \p{\lim_{n\to\infty} b_n}^2 = b^2,

and by Theorem 9.2, limn4an=4a\displaystyle\lim_{n\to\infty} 4a_n = 4a. Thus, by Theorem 9.3,

limn(an3+4an)=(limnan3)+(limn4an)=a3+4a,andlimn(bn2+1)=(limnbn2)+(limn1)=b2+1.\begin{aligned} \lim_{n\to\infty} \p{a_n^3 + 4a_n} = \p{\lim_{n\to\infty} a_n^3} + \p{\lim_{n\to\infty} 4a_n} &= a^3 + 4a, \quad\text{and} \\ \lim_{n\to\infty} \p{b_n^2 + 1} = \p{\lim_{n\to\infty} b_n^2} + \p{\lim_{n\to\infty} 1} &= b^2 + 1. \end{aligned}

Since b2+11b^2 + 1 \geq 1, we know that b2+10b^2 + 1 \neq 0, so now we can apply Theorem 9.6 to get

limnsn=limn(an3+4an)limn(bn2+1)=a3+4ab2+1.\lim_{n\to\infty} s_n = \frac{\lim_{n\to\infty} \p{a_n^3 + 4a_n}}{\lim_{n\to\infty} \p{b_n^2 + 1}} = \frac{a^3 + 4a}{b^2 + 1}.

Continuity

Introduction

Definition

Suppose f ⁣:DR\func{f}{D}{\R} is a function. Then ff is continuous at x0Dx_0 \in D if one of the following holds:

  1. If {xn}n\set{x_n}_n is a sequence in DD converging to x0x_0, then {f(xn)}n\set{f\p{x_n}}_n converges to f(x0)f\p{x_0}.
  2. For any ε>0\epsilon > 0, there exists δ>0\delta > 0 such that whenever yDy \in D satisfies x0y<δ\abs{x_0 - y} < \delta, then f(x0)f(y)<ε\abs{f\p{x_0} - f\p{y}} < \epsilon.
Remark.

From lecture, you already know that (i) and (ii) are equivalent definitions, which means we can use whichever one's more convenient for us. As a rule of thumb, (i) is easier to use to prove something is discontinuous, and (ii) is easier to use to prove something is continuous.

Example 1.
(17.10(b))

Prove that

f(x)={sin1xif x>0,0if x=0f\p{x} = \begin{cases} \sin{\frac{1}{x}} & \text{if } x > 0, \\ 0 & \text{if } x = 0 \end{cases}

is not continuous at x0=0x_0 = 0.

Solution.

Following the remark above, we're going to want to use (i). To prove discontinuity, we just need to find a sequence {xn}n\set{x_n}_n in [0,)\pco{0, \infty} such that {xn}n\set{x_n}_n converges to 00, but {f(xn)}n\set{f\p{x_n}}_n does not converge to f(0)=0f\p{0} = 0. Since sinθ=1\sin{\theta} = 1 at π2+2πn\frac{\pi}{2} + 2\pi n for all nNn \in \N, we can set

1xn=π2+2πn    xn=1π2+2πn.\frac{1}{x_n} = \frac{\pi}{2} + 2\pi n \implies x_n = \frac{1}{\frac{\pi}{2} + 2\pi n}.

Then xn0x_n \to 0 as nn \to \infty, but f(xn)=1f\p{x_n} = 1 for all nNn \in \N, so ff is not continuous at x0=0x_0 = 0.

Example 2.
(17.4)

Prove that x\sqrt{x} is continuous on [0,)\pco{0, \infty}.

Solution.

We're going to use definition (ii) here, which means that we want to show an inequality of the form

x0y<ε,\abs{\sqrt{x_0} - \sqrt{y}} < \epsilon,

so let's take a look at the left-hand side:

x0yx0+yx0+y=x0yx0+yx0yx0.\abs{\sqrt{x_0} - \sqrt{y}} \cdot \abs{\frac{\sqrt{x_0} + \sqrt{y}}{\sqrt{x_0} + \sqrt{y}}} = \frac{\abs{x_0 - y}}{\sqrt{x_0} + \sqrt{y}} \leq \frac{\abs{x_0 - y}}{\sqrt{x_0}}.

This inequality is okay except when x0=0x_0 = 0, which tells me that I want to deal with this case separately.

Let ε>0\epsilon > 0. If x0=0x_0 = 0, then if y<δ\abs{y} < \delta,

y=y<δ.\abs{\sqrt{y}} = \sqrt{y} < \sqrt{\delta}.

So, δ=ε2\delta = \epsilon^2 gives y<ε\abs{\sqrt{y}} < \epsilon, so x\sqrt{x} is continuous at x0=0x_0 = 0.

If x00x_0 \neq 0, then we can use the inequality above: if x0y<δ\abs{x_0 - y} < \delta, then

x0yx0yx0<δx0.\abs{\sqrt{x_0} - \sqrt{y}} \leq \frac{\abs{x_0 - y}}{\sqrt{x_0}} < \frac{\delta}{\sqrt{x_0}}.

Thus, δ=εx0\delta = \epsilon\sqrt{x_0} works: x0y<εx0x0=ε\abs{\sqrt{x_0} - \sqrt{y}} < \epsilon\frac{\sqrt{x_0}}{\sqrt{x_0}} = \epsilon. Thus, x\sqrt{x} is continuous on all of [0,)\pco{0, \infty}.

Another consequence of the sequence definition of continuity is that facts about sequences translate directly into facts about continuity.

Theorem

Suppose f ⁣:DR\func{f}{D}{\R} and g ⁣:DR\func{g}{D}{\R} are continuous functions. Then:

  1. f+gf + g is continuous.
  2. fgfg is continuous.
  3. fg\frac{f}{g} is continuous at all points x0Dx_0 \in D such that g(x0)0g\p{x_0} \neq 0.
Proof.
(sketch)

Continuity at a point x0Dx_0 \in D basically boils down to sequences. For f+gf + g, if {xn}n\set{x_n}_n is a sequence in DD that converges to x0x_0, then we need to show that {(f+g)(xn)}n\set{\p{f + g}\p{x_n}}_n converges to (f+g)(x0)\p{f + g}\p{x_0}, but this boils down to the limit laws for sequences. \square

This theorem tells us that sequences help us with proving continuity. The converse is also true: we can use continuity to help us calculate limits.

Example 3.

Calculate limne1/n\displaystyle\lim_{n\to\infty} e^{1/n}.

Solution.

exe^x is continuous (which is a fact we'll just take for granted), and 1n\frac{1}{n} converges to 00. Thus, by the sequence definition of continuity,

limne1/n=elimn(1/n)=e0=1.\lim_{n\to\infty} e^{1/n} = e^{\lim_{n\to\infty} \p{1/n}} = e^0 = 1.

To end this section, here's an important connection between density and continuity:

Proposition

Let f ⁣:RR\func{f}{\R}{\R} and g ⁣:RR\func{g}{\R}{\R} be continuous functions such that f(q)=g(q)f\p{q} = g\p{q} for all qQq \in \Q. Then f=gf = g everywhere.

Remark.

The idea behind this fact is that a continuous function will "fill in gaps" through limits. For example, let's say I give you all the information about a function f(x)f\p{x} except at a single point:

Because of continuity, there's only one way you can fill in the gap. In other words, information about a continuous function except at one point is the same thing as information about the continuous function everywhere.

The proposition says something a bit stronger: information about a continuous function at the rational numbers is the same as information about the function on the whole real line.

Proof.

To prove the statement, we just need to use density of Q\Q. Let x0Rx_0 \in \R, so by density of Q\Q, there exists a sequence of rational numbers {qn}n\set{q_n}_n which converge to x0x_0. Since we know that ff and gg agree at every rational, we have f(qn)=g(qn)f\p{q_n} = g\p{q_n} for all nn. Then by the sequence definition of continuity,

f(x0)=limnf(qn)=limng(qn)=g(x0).f\p{x_0} = \lim_{n\to\infty} f\p{q_n} = \lim_{n\to\infty} g\p{q_n} = g\p{x_0}.

Since x0x_0 was arbitrary, this means f=gf = g everywhere on R\R, which was what we wanted to show. \square

True or False

Example 4.

True or false: f ⁣:R{0}R\func{f}{\R \setminus \set{0}}{\R} defined by f(x)=1xf\p{x} = \frac{1}{x} is continuous.

Solution.

True. Continuity of a function only depends on points in the domain. x0=0x_0 = 0 is a problematic point for ff, but 00 is outside the domain of ff, so we don't care what happens there.

Example 5.

True or false: If f+gf + g is continuous, then ff or gg is must be continuous at at least one point.

Solution.

False. For example, let

f(x)={1if xQ,0otherwise.f\p{x} = \begin{cases} 1 & \text{if } x \in \Q, \\ 0 & \text{otherwise}. \end{cases}

Then if g=fg = -f, we get f+g=0f + g = 0 which is continuous everywhere, but ff and gg are both discontinuous everywhere. This is because Q\Q and RQ\R \setminus \Q are both dense in R\R, so given any x0Rx_0 \in \R, there exist a sequence {qn}n\set{q_n}_n of rational numbers and a sequence {rn}n\set{r_n}_n of irrational numbers which both converge to x0x_0. But

f(qn)=1andf(rn)=0for all nN,f\p{q_n} = 1 \quad\text{and}\quad f\p{r_n} = 0 \quad\text{for all } n \in \N,

so by the sequence definition of continuity, ff is discontinuous everywhere.

Example 6.

True or false: If f+gf + g is continuous and ff is continuous, then gg is continuous.

Solution.

True. You can write

g=(f+g)continuous+(f)continuous,g = \underbrace{\p{f + g}}_{\text{continuous}} + \underbrace{\p{-f}}_{\text{continuous}},

so gg is a sum of continuous functions, hence continuous.