Chapter 5 Functions on Metric Spaces and Continuity

When we studied real-valued functions of a real variable in calculus, the techniques and theory built on properties of continuity, differentiability, and integrability. All of these concepts are dened using the precise idea of a limit. In this chapter, we want to look at functions on metric spaces. In particular, we want to see how mapping metric spaces to metric spaces relates to properties of subsets of the metric spaces.

5.1 Limits of Functions

Recall the denitions of limit and continuity of real-valued functions of a real variable. Denition 5.1.1 Suppose that f is a real-valued function of a real variable, p + U, and there is an interval I containing p which, except possibly for p is in the domain of f . Then the limit of f as x approaches p is L if and only if 1   0 " 2= =   =
x p

0 F 1x  0  x  p   = "  f x   L    . L which is read as the limit of f of x as x

In this case, we write lim f  x  approaches p is equal to L.

Denition 5.1.2 Suppose that f is a real-valued function of a real variable and p + dom  f . Then f is continuous at p if and only if lim f  x  f  p .
xp

183

184

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

These are more or less the way limit of a function and continuity of a function at a point were dened at the time of your rst encounter with them. With our new terminology, we can relax some of what goes into the denition of limit. Instead of going for an interval (with possibly a point missing), we can specify that the point p be a limit point of the domain of f and then insert that we are only looking at the real numbers that are both in the domain of the function and in the open interval. This leads us to the following variation. Denition 5.1.3 Suppose that f is a real-valued function of a real variable, dom  f  A, and p + A) (i.e., p is a limit point of the domain of f ). Then the limit of f as x approaches p is L if and only if 0 " 2= =d  0 1   e 1x   x + A F 0  x  p   = "  f x   L     b Example 5.1.4 Use the denition to prove that lim 2x 2
x 3

4x

31.

Before we offer a proof, well illustrate some expanded scratch work that leads to the information needed in order to offer a proof. We want to show that, corresponding to 0 we can nd a = 0 such that 0  x  3  = " nb 2 n c each n 2x n 4x 1  31  . The easiest way to do this is to come up with a = that is a function of . Note that nr n n n s n n n n n 2x 2 4x 1  31n n2x 2 4x  30n 2 x  3 x 5 . The x  3 is good news because it is ours to make as small as we choose. But if we restrict x  3 there is a corresponding restriction on x 5 to take care of this part we will put a cap on = which will lead to simpler expressions. Suppose that we place a 1st restriction on = of requiring that = n 1. If = n 1, then 0  x  3  = n 1 " x 5  x  3 8 n x  3 8  9. Now n nr s n n 2 4 x 1  31n 2 x  3 x 5  2  =  9 n n 2x whenever = n . To get both bounds to be in effect we will take = 18 This concludes that expanded scratch work. Q R Proof. For 0, let = max 1 . Then 18 0   x  3  = n 1 "  x 5 x  3 8 n x  3 Q max 1 R .

18

89

5.1. LIMITS OF FUNCTIONS and nr n n 2x 2 n n 1  31n s

185

4x

2 x  3 x

5  2  =  9 n 18 

18

Since 0 was arbitrary, we conclude that, for every 0, there exists a = Q R nb 2 n c min 1 0, such that 0  x  3  = " n 2x 4x 1  31n   i.e., b 18 c lim 2x 2 4x 1 31.
x 3

b Excursion 5.1.5 Use the denition to prove that lim x 2

x 1

5x

6.

Space for scratch work.

A Proof.

***For this one, the = that you dene will depend on the nature of the rst restriction that you placed on = in order to obtain a niceQupper R bound on x 6 if you chose = n 1 as your rst restriction, then = min 1 would have been what worked 8 in the proof that was offered.*** You want to be careful not to blindly take = n 1 as the rst restriction. For 1 example, if you are looking at the greatest integer function as x  , you would 2 1 need to make sure that = never exceeded in order to stay away from the nearest 2 1 jumps if you have a rational function for which is a zero of the denominator 2 1 1 and you are looking at the limit as x  , then you couldnt let = be as great at 4 4

186

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

1 so you might try taking = n as a rst restriction. Our next example takes such a 6 consideration into account. Example 5.1.6 Use the denition to prove that lim Space for scratch work. x2 x 1 2 x 3 1 4.

Proof. For that x and 2 x 1 n n 2n nx

0, let =

} 1 2 min  . From 0  x 4 25 1 n n 2n n x 6  x 1 6 1 4

1  = n

1 , we have 4

x n 1n n 2n

25 4 t u 1 2 4 t u

n n n 1n nx 1  n o 2 n 2n

n 1n 1  n 2n

1 . 2

Furthermore, nt 2 n x n n 2x n u n 3  4n n 1 n 2 n n nx 8 x 7 n n n 2x 1 n 25  = 25 2 n  2 2 25 25 = x 1 x 7 4  1 2x 1 2 . |

} 1 2 Since 0 was arbitrary, for every 0 there exists a = min  4 25 nt 2 n u n x n 3 n 0 such that 0  x 1  = implies that n n 2 x 1  4n   that is, x2 x 1 2x lim 3 1 4.

5.1. LIMITS OF FUNCTIONS

187

In Euclidean U space, the metric is realized as the absolute value of the difference. Letting d denote this metric allows us to restate the denition of lim f x 
x p

L as 1   d e 1x  x + A F 0  d x  p   = " d  f  x   L    . 0 " 2= =  0

Of course, at this point we havent gained much this form doesnt look particularly better than the one with which we started. On the other hand, it gets us nearer to where we want to go which is to the limit of a function that is from one metric space to anotherneither of which is U1 . As a rst step, lets look at the denition when the function is from an arbitrary metric space into U1 . Again we let d denote the Euclidean 1metric. Denition 5.1.7 Suppose that A is a subset of a metric space  S  d S  and that f is a function with domain A and range contained in U1  i.e., f : A  U1 . then f tends to L as x tends to p through points of A if and only if (i) p is a limit point of A, and (ii) 1 0 2= =  0 1x  x + A F 0  d S  x  p  = " d  f x   L   . In this case, we write f  x   L as x  p for x + A, or f x   L as x  p, or
A x p
x+A

lim f  x 

L

Example 5.1.8 Let f : F  U be given by f z 

z 3 i
z +F

Re z . Prove that

lim f z 

3

Space for scratch work.

188

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

For this one, we will make use of the fact that for any complex number ? , Re ?  n ? .

Proof. For  f z   3 Since

0, let =

. Then 0  z  3 Re z  3
z 3 i
z +F

i   = i  n z  3 3.

implies that i   .

Re z   3

0 was arbitrary, we conclude that lim f z 

Remark 5.1.9 Notice that, in the denition of lim f  x 

x p
x+A

L, there is neither a

requirement that f be dened at p nor an expectation that p be an element of A. Also, while it isnt indicated, the = 0 that is sought may be dependent on p. Finally we want to make the transition to functions from one arbitrary metric space to another. Denition 5.1.10 Suppose that A is a subset of a metric space  S  d S  and that f is a function with domain A and range contained in a metric space  X  d X  i.e., f : A  X. Then f tends to L as x tends to p through points of A if and only if (i) p is a limit point of A, and (ii) 1 0 2= =  0 1x  x + A F 0  d S  x  p  = " d X  f  x   L   . In this case, we write f  x   L as x  p for x + A, or f  x   L as x  p, or
A x p
x+A

lim f  x 

L

5.1. LIMITS OF FUNCTIONS

189

Example 5.1.11 For p + U1 , let f  p  2 p 1, p 2 . Then f : U1  U2 . Use the denition of limit to from that lim f  p  3 1 with respect to the Euclidean
p 1

metrics on each space. Space for scratch work.

Proof. For = n 1 implies that

0, let = p 1 T 4

} | min 1 T . Then 0  dU  p  1 13  p  1 2 n  p  1 T 4 23

 p  1 

and

p

12 

T 13.

Hence, for 0  dU  p  1

 p  1  = , T b c2 p2  1 dU2  f  p  3 1 2 p 1  32 T T  p  1 4  p 12  =  13 n .

p 1

Since

0 was arbitrary, we conclude that lim 2 p

1, p 2 

3 1.

Remark 5.1.12 With few exceptions our limit theorems for functions of real-valued functions of a real variable that involved basic combinations of functions have direct, straightforward analogs to functions on an arbitrary metric spaces. Things can get more difcult when we try for generalizations of results that involved comparing function values. For the next couple of excursions, you are just being asked to practice translating results from one setting to our new one. Excursion 5.1.13 Let A be a subset of a metric space S and suppose that f : A  U1 is given. If f  L as p  p0 in A and f  M as p  p0 in A

190

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

prove that L M. After reading the following proof for the case of real-valued functions of a real variable, use the space provided to write a proof for the new setting.

Proof. We want to prove that, if f  L as x  a and f  M as x  a, then 1 L M . For L / M , let > 2   L  M . By the denition of limit, there exists positive numbers =1 and =2 such that 0   x  a   =1 implies  f  x   L   > and 0  x  a   =2 implies  f  x   M   > . Choose x0 + U such that 0  x0  a   min
=1  =2 . Then  L  M  n  L  f  x0   M  f  x0   2> which contradicts the trichotomy law.

Excursion 5.1.14 Let f and g be real-valued functions with domain A, a subset of a metric space  S  d . If lim f  p  L and lim g  p  M, then
p p0
p+ A

p  p0
p+ A

p p0
p+ A

lim  f

g  p 

M.

After reading the following proof for the case of real-valued functions of a real variable, use the space provided to write a proof for the new setting.

Proof. We want to show that, if lim f  x  L and lim g x  M, then x a x a lim  f g x  L M . Let > 0 be given. Then there exists positive numbers =1 x a and =2 such that 0  x  a   =1 implies  f  x   L   >2 and 0  x  a   =2 implies g x   M   >2. For = min
=1  =2 , 0  x  a   = implies that

5.1. LIMITS OF FUNCTIONS  f g x    L M  n  f x   L  g x   M   > .

191

Theorem 4.1.17 gave us a characterization of limit points in terms of limits of sequences. This leads nicely to a characterization of limits of functions in terms of behavior on convergent sequences. Theorem 5.1.15 (Sequences Characterization for Limits of Functions) Suppose that  X  d X  and Y dY  are metric spaces, E t X, f : E  Y and p is a limit point of E. Then lim f  x  q if and only if
xp
x +E

1
pn 

Kr

pn  t E F 1n   pn /

p F lim pn
n *

s p " lim f  pn 
n *

L q .

Excursion 5.1.16 Fill in what is missing in order to complete the following proof of the theorem. Proof. Let X , Y , E , f , and p be as described in the introduction to Theorem 5.1.15. Suppose that lim f  x  q . Since p is a limit point of E , by Theorem
x p
x+E

, there exists a sequence

pn  of elements in E such that pn / n + M, and exists =
2 1

p for all q , there .

3

. For

0, because lim f  x 
x p
x+E

0 such that 0  d X  x  p   = and x + E implies that

From lim pn p and pn / p, we also know that there exists a positive integer M n * such that n M implies that . Thus, it follows that
4

192

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

dY  f  pn   q   for all n M . Since 0 was arbitrary, we conclude that lim f  pn  q . Finally, because
pn  t E was arbitrary, we have that
n *

" lim f  pn 
5 n *

1
pn 

q . q.

We will give a proof by contrapositive of the converse. Suppose that lim f  x  /

x p
x+E

Then there exists a positive real number such that corresponding to each positive real number = there is a point x= + E for which 0  d X  x=  p  = and 1 dY  f  x=   q  o . In particular, for each n + M, corresponding to there is a point n pn + E such that and dY  f  pn   q  o . Hence, lim f  pn  / q . Thus, there exists a sequence
pn  t E such that lim pn
n * 6 n *

p and
7

 L q

i.e., 2
pn  Kr
pn  t E F 1n   pn / p F lim pn
n *

s p F lim f  pn  /
n *

which is equivalent to Kr  1
pn 
pn  t E F 1n   pn /
x p
x +E

p  F lim pn
n *

s p " lim f  pn 
n *

L q .

Therefore, we have shown that lim f x  /  1

pn  Since the
8 

q implies that s p " lim f  pn 

n *

Kr

pn  t E F 1n   pn /

p  F lim pn
n *

L q .

is logically equivalent to

the converse, this concludes the proof. ***Acceptable responses are: (1) 4.1.17,(2) lim pn p, (3) dY  f  x   q   , n * (4) 0  d X  pn  p   = , (5)
pn  t E F 1n   pn / p F lim pn p, (6) n * 1 0  d X  pn  p   , (7) lim f  pn  / q , (8) contrapositive.*** n * n The following result is an immediate consequence of the theorem and Lemma 4.1.7.

5.1. LIMITS OF FUNCTIONS Corollary 5.1.17 Limits of functions on metric spaces are unique.

193

Remark 5.1.18 In view of Theorem 5.1.15, functions from metric spaces into subsets of the complex numbers will satisfy the limits of combinations properties of sequences of complex numbers that were given in Theorem 4.3.2. For completeness, we state it as a separate theorem. Theorem 5.1.19 Suppose that  X  d X  is a metric space, E t X, p is a limit point of E, f : E  F, g : E  F, lim f x  A, and lim g x  B. Then
x p
x +E

x p
x+E

(a) lim  f
xp
x +E

g  x 

A AB

(b) lim  f g   x 
xp
x +E

(c) lim
x +E

xp

f x  g

A whenever B / B

0.

While these statements are an immediate consequence of Theorem 4.3.2 and Theorem 5.1.15 completing the following excursions can help you to learn the approaches to proof. Each proof offered is independent of Theorems 4.3.2 and Theorem 5.1.15. Excursion 5.1.20 Fill in what is missing to complete a proof of Theorem 5.1.19(a). Proof. Suppose 0 is given. Because lim f  x 
xp
x +E

A, there exists a positive

real =1 such that x + E and 0  d X  x  p   =1 implies that  f  x   A  . 2 Since , there exists a positive real number =2 such that x + E and
1

0  d X  x  p   =2 implies that g  x   B  

. Let = . It 2 2 follows from the triangular inequality that, if x + E and 0  d X  x  p   = , then n n n n n n  f g  x    A B  n f  x   A n n n 3 n
4


5

194 Since

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY 0 was arbitrary, we conclude that lim  f
xp
x +E

g  x 

B as claimed.

***Acceptable responses are: (1) lim g x 

x p
x +E

B , (2) min
=1  =2 , (3) g x   B ,

(4)  f x   A

g x   B , (5) ***

Excursion 5.1.21 Fill in what is missing to complete a proof of Theorem 5.1.19(b).

Proof. Because lim f  x 

x p
x+E

A, there exist a positive real number =1 such that

x + E and 0  d X  x  p   =1 implies that  f  x   A  1 i.e.,  f x  A  1. Hence,  f  x   1  A for all x + E such that 0  d X  x  p   =1 . Suppose that 0 is given. If B 0, then lim g  x  0 yields the
x p
x+E

existence of a positive real number =2 such that x + E and 0  d X  x  p  =2 implies that g  x   Then for = `
1

, we have that

 f g  x 

 f x  g x   1

 A 
2

Hence, lim  f g  x 
x p
x+E

AB

0. Next we suppose that B /

0. Then there exists a

positive real numbers =3 and =4 for which  f  x   A  2 1  A x + E . Now let =

and g  x   B   2 B whenever 0  d X  x  p   =3 and 0  d X x  p  =4 , respectively, for min

=1  =3  =4 . It follows that if x + E and 0  d X  x  p  

5.1. LIMITS OF FUNCTIONS = , then  f g   x   AB  n n n n f x  g x   n n n n  AB n n n n n n n .

7 8 xp
x +E

195

3

4 5

n  f  x  g  x   B   1  Again, since 0 was arbitrary, we conclude that lim  f g   x   A g  x   B  n n n B n n

6

AB as

needed. ***Acceptable responses include: (1) min

=1  =2 , (2) 1  A1 , (3) f x  B , B (4) f  x  B , (5)  B   f x   A, (6)  f x   A, (7) 1  A 2 1  A 2 B (8) .*** Excursion 5.1.22 Fill in what is missing to complete a proof of Theorem 5.1.19(c). Proof. In view of Theorem 5.1.19(b), it will sufce to prove that, under the 1 1 given hypotheses, lim . First, we will show that, for B / 0, the modulus x  p g x  B of g is bounded away from zero. Since  B  a positive real number =1
x+E

0 and lim g x 
x p
x+E

B , there exists implies that

0 such that x + E and

B g  x   B   . It follows from the (other) 2 if x + E and 0  d X  x  p   =1 , then n n g x   g x   B  B  o ng  x   B  B . 2

1

that,
2 

n n n

196

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY 0 is given. Then  B 2 2 0 and lim g  x 
xp
x+E

Suppose that

B yields the

 B 2 existence of a positive real number =2 such that g x   B   whenever 2 min

=1  =2 . Then for x + E and x + E and 0  d X x  p  =2 . Let = 0  d X x  p   = we have that n n n 1 1n n n g  x   B   . n g x   B n  B  g  x  3  4 Since 0 was arbitrary, we conclude that
5

. ,
6

Finally, letting h x  f x  x p g lim

x+E

1 , by Theorem g x  lim f  x 
x+E

xp

.
7 8

***Acceptable responses are: (1) 0  d X  x  p  =1 , (2) triangular inequality (3)  B 2 1 1 1 t u , (4) (5) lim , (6) 5.1.19(b), (7) h  x , (8) A  .*** x  p g x  B B B x+E 2 B 2 From Lemma 4.3.1, it follows that the limit of the sum and the limit of the product parts of Theorem 5.1.19 carry over to the sum and inner product of functions from metric spaces to Euclidean k  space. Theorem 5.1.23 Suppose that X is a metric space, E t X, p is a limit point of E, f : E  Uk , g : E  Uk , lim f  x  A, and lim g  x  B. Then
x p
x+E

x p
x+E

(a) lim f
xp
x +E

g x 

B and

(b) lim f  g  x 
xp
x +E

AB

In the set-up of Theorem 5.1.23, note that f g : E  Uk while fg : E  U.

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES

197

5.2 Continuous Functions on Metric Spaces

Recall that in the case of real-valued functions of a real variable getting from the general idea of a functions having limits to being continuous simply added the property that the values approached are actually the values that are achieved. There is nothing about that transition that was tied to the properties of the reals. Consequently, the denition of continuous functions on arbitrary metric spaces should come as no surprise. On the other hand, an extra adjustment is needed to allow for the fact that we can consider functions dened at isolated points of subsets of metric spaces. Denition 5.2.1 Suppose that  X  d X  and Y dY  are metric spaces, E t X, f : E  Y and p + E. Then f is continuous at p if and only if d e 0 2= =   0 1x  x + E F d X  x  p   = " dY  f  x   f  p  . 1 Theorem 5.2.2 Suppose that  X  d X  and Y dY  are metric spaces, E t X, f : E  Y and p + E and p is a limit point of E. Then f is continuous at p if and only if lim f  x  f  p
xp
x+E

Denition 5.2.3 Suppose that  X  d X  and Y  dY  are metric spaces, E t X and f : E  Y . Then f is continuous on E if and only if f is continuous at each p + E. Remark 5.2.4 The property that was added in order to get the characterization that is given in Theorem 5.2.2 was the need for the point to be a limit point. The denition of continuity at a point is satised for isolated points of E because each isolated point p has the property that there is a neighborhood of p, N=`  p , for which E D N= `  p 
p  since p + dom  f  and d X  p  p  dY  f  p  f  p  0, we automatically have that 1x   x + E F d X x  p  = " dY  f x   f  p   for any 0 and any positive real number = such that =  = ` . Remark 5.2.5 It follows immediately from our limit theorems concerning the algebraic manipulations of functions for which the limits exist, the all real-valued polynomials in k real variables are continuous in Uk . bb cc Remark 5.2.6 Because f 1 2 p 1 p2 : U1  3 1 for the f  p U2 that was given in Example 5.1.11, our work for the example allows us to claim that f is continuous at p 1.

198

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

Theorem 5.1.15 is not practical for use to show that a specic function is continuous it is a useful tool for proving some general results about continuous functions on metric spaces and can be a nice way to show that a given function is not continuous. Example 5.2.7 Prove that the function f : U  U  U given by f  x  y   xy ! , for x  y  / 0 0  3 x y3 is not continuous at 0 0. !  0 , foru x y 0 t 1 1 Let pn  . Then
pn * n 1 converges to 0 0, but n n t ut u 1 1 n n n lim f  pn  lim t u3 t u3 lim * / 0. n * n * 1 n * 2 1 n n

Hence, by the Sequences Characterization for Limits of Functions, we conclude that the given f is not continuous at 0 0. Example 5.2.8  Use2 the denition to prove that f : U  U  U given by x y ! !  2 , for x  y  / 0 0 2 x y is continuous at 0 0. f  x  y  ! !  0 , for x y 0 We need to show that lim f x  y  0. Because the function is dened in two parts, it is necessary to appeal to the denition. For T 0  dUU  x  y   0 0 x 2 y2  = implies that  f x  y   0 c n 2 n b 2 n x y n x y2  y n n n x 2 y2 n n x 2 y2  y T y2 n T x2 y2  .
x  y 00

0, let =

. Then

Because 0 was arbitrary, we conclude that lim f x  y  0 f 0 0. Hence, f is continuous at 0 0.
x  y 00

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES

199

It follows from the denition and Theorem 5.1.19 that continuity is transmitted to sums, products, and quotients when the ranges of our functions are subsets of the complex eld. For completeness, the general result is stated in the following theorem. Theorem 5.2.9 If f and g are complex valued functions that are continuous on f a metric space X, then f g and f g are continuous on X. Furthermore, is g continuous on X 
p + X : g  p  0. From Lemma 4.3.1, it follows immediately that functions from arbitrary metric spaces to Euclidean k -space are continuous if and only if they are continuous by coordinate. Furthermore, Theorem 5.1.23 tells us that continuity is transmitted to sums and inner products. Theorem 5.2.10 (a) Let f 1  f 2   f k be real valued functions on a metric space X, and F : X  Uk be dened by F x   f 1  x   f 2  x    f k x . Then F is continuous if and only if f j is continuous for each j  1 n j n k. (b) If f and g are continuous functions from a metric space X into Uk , then f and f  g are continuous on X. g

The other combination of functions that we wish to examine on arbitrary metric spaces is that of composition. If X , Y , and Z are metric spaces, E t X , f : E  Y , and g : f  E   Z , then the composition of f and g, denoted by g i f , is dened by g  f x  for each x + E . The following theorem tells us that continuity is transmitted through composition. Theorem 5.2.11 Suppose that X, Y , and Z are metric spaces, E t X, f : E  Y , and g : f  E   Z. If f is continuous at p + E and g is continuous at f  p , then the composition g i f is continuous at p + E. Space for scratch work.

Proof. Suppose that  X  d X , Y  dY , and  Z  d Z  are metric spaces, E t X , f : E  Y , g : f  E   Z , f is continuous at p + E , and g is continuous at f  p . Let 0 be given. Since g is continuous at f  p , there exists

200

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

a positive real number =1 such that d Z g  y   g  f  p   for any y + f  E  such that dY  y  f  p  =1 . From f being continuous at p + E and =1 being a positive real number, we deduce the existence of another positive real number = such that x + E and d X  x  p   = implies that dY  f  x   f  p   =1 . Substituting f  x  for y , we have that x + E and d X  x  p  = implies that dY  f  x   f  p   =1 which further implies that d Z  g  f  x   g  f  p   . That is, d Z  g i f  x   g i f   p  for any x + E for which d X  x  p   = . Therefore, g i f is continuous at p .

Remark 5.2.12 The with respect to a set  distinction can be an important one to  1 , for x rational note. For example, the function f x  is continuous  0 , for x irrational with respect to the rationals and it is continuous with respect to the irrationals. However, it is not continuous on U1 .

5.2.1

A Characterization of Continuity

Because continuity is dened in terms of proximity, it can be helpful to rewrite the denition in terms of neighborhoods. Recall that, for  X  d X , p + X , and = 0, N=  p  N  f  p 
x + X : d X  x  p  =  . 0,
y + Y : dY  y  f  p    . For a metric space Y dY , f : X  Y and

Hence, for metric spaces  X  d X  and Y dY , E t X , f : E  Y and p + E , f is continuous at p if and only if d e 0 2= =   0 f  N=  p  D E  t N  f  p . 1 Because neighborhoods are used to dene open sets, the neighborhood formulation for the denition of continuity of a function points us in the direction of the following theorem.

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES

201

Theorem 5.2.13 (Open Set Characterization of Continuous Functions) Let f be a mapping on a metric space  X  d X  into a metric space Y dY . Then f is continuous on X if and only if for every open set V in Y , the set f 1  V  is open in X. Space for scratch work.

Excursion 5.2.14 Fill in what is missing in order to complete the following proof of the theorem. Proof. Let f be a mapping from a metric space  X  d X  into a metric space Y  dY . Suppose that f is continuous on X , V is an open set in Y , and p0 +  1 f  V . Since V is open and f  p0  + V , we can choose 0 such that N  f  p0  t V from which it follows that t f 1  V  .
1

Because f is continuous at p0 + X , corresponding to such that f  N=  p0  t N  f  p0  which implies that N=  p0  t

2 

0, there exists a =

0,

From the transitivity of subset, we concluded that N=  p0  t f 1  V . Hence, p0 is an interior point of f 1  V . Since p0 was arbitrary, we conclude that each p + f 1  V  is an interior point. Therefore, f 1 V  is open. To prove the converse, suppose that the inverse image of every open set in Y is open in X . Let p be an element in X and 0 be given. Now the neighborhood N  f  p is open in Y . Consequently, is open in X .
3 

202

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY , there exists a positive real number = such
3

Since p is an element of that N=  p t f 1  N  f  p  i.e.,

t N  f  p . Since
4 5

was arbitrary, we conclude that lim f  x 

xp

. Finally, because p was an as needed.

6

arbitrary point in X , it follows that f

***Acceptable responses are: (1) f 1  N  f  p0 , (2) f 1  N  f  p0 , (3) f 1  N  f  p, (4) f  N=  p , (5) f  p, (6) is continuous on X .*** Excursion 5.2.15 Suppose that f is a mapping on spacec X  d X  into a d a emetric b  c  1 c metric space Y  dY  and E t X. Prove that f E f 1 [E ] .

The following corollary follows immediately from the Open Set Characterization for Continuity, Excursion 5.2.15, and the fact that a set is closed if and only if its complement is open. Use the space provided after the statement to convince yourself of the truth of the given statement. Corollary 5.2.16 A mapping f of a metric space X into a metric space Y is continuous if and only if f 1 C  is closed in X for every closed set C in Y .

Remark 5.2.17 We have stated results in terms of open sets in the full metric space. We could also discuss functions restricted to subsets of metric spaces and then the characterization would be in terms of relative openness. Recall that given two sets X and Y and f : X  Y , the corresponding set induced functions satisfy the following properties for C j t X and D j t Y , j 1 2:

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES  f 1 [ D1 D D2 ]  f 1 [ D1 C D2 ] f 1 [ D1 ] D f 1 [ D2 ], f 1 [ D1 ] C f 1 [ D2 ],

203

 f [C1 D C2 ] t f [C1 ] D f [C2 ], and  f [C1 C C2 ] f [C1 ] C f [C2 ]

Because subsets being open to subsets of metric spaces in characterized by their realization as intersections with open subsets of the parent metric spaces, our neighborhoods characterization tells us that we loose nothing by looking at restrictions of given functions to the subsets that we wish to consider rather that stating things in terms of relative openness.

5.2.2

Continuity and Compactness

Theorem 5.2.18 If f is a continuous function from a compact metric space X to a metric space Y , then f  X  is compact. Excursion 5.2.19 Fill in what is missing to complete the following proof of Theorem 5.2.18. Space for scratch work. Proof. Suppose that f is a continuous function from a compact metric space X to a metric space Y and J
G : : : +  is an open cover for f  X . Then G : is open in Y for each : +  and .
1

From the Open Set Characterization of Functions, f 1  G

2  :

is
3

for each : + . 6
: +

Since f : X  f  X  and f  X  t that X f 1  f  X  t f 1 >

: +

G : , we have

G:
4

204

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY j k f 1 G :  : : +  is an for X . Since X is , there is a nite subcollection of I , k 1 2  n , that covers X  i.e., Xt
n > j 1

Hence, I

5

b6 c f 1 G : j : j

b c f 1 G : j .

It follows that n > b c f X t f f 1 G : j

j 1

n > j 1 7

n > j 1

G: j .

j k Therefore, G : j : j 1 2  n is a nite subcollection of J that covers f  X . Since J was arbitrary, every  i.e., f  X  is
9 8

Remark 5.2.20 Just to stress the point, in view of our denition of relative compactness the result just stated is also telling us that the continuous image of any compact subset of a metric space is a compact subset in the image. Denition 5.2.21 For a set E, a function f : E  Uk is said to be bounded if and only if 2 M   M + UF 1x   x + E " f  x  n M  .

When we add compactness to domain in the metric space, we get some nice analogs. Theorem 5.2.22 (Boundedness Theorem) Let A be a compact subset of a metric space  S  d  and suppose that f : A  Uk is continuous. Then f  A is closed and bounded. In particular, f is bounded.

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES

205

Excursion 5.2.23 Fill in the blanks to complete the following proof of the Boundedness Theorem. Proof. By the , we know that compactness in Uk for any k + M is equivalent to being closed and bounded. Hence, from Theorem 5.2.18, if f : A  Uk where A is a compact metric space, then f  A is compact. But f  A t and compact yields that f  A is . In particular, f  A is bounded as claimed in the Boundedness Theorem. ***Expected responses are: (1) Heine-Borel Theorem, (2) Uk , and (3) closed and bounded.*** Theorem 5.2.24 (Extreme Value Theorem) Suppose that f is a continuous function from a compact subset A of a metric space S into U1 , M sup f  p 
p+ A 2 3 1

and

p+ A

inf f  p  . M and f ) m.

Then there exist points u and ) in A such that f u 

Proof. From Theorem 5.2.18 and the Heine-Borel Theorem, f  A t U and f continuous implies that f  A is closed and bounded. The Least Upper and Greatest Lower Bound Properties for the reals yields the existence of nite real numbers M and m such that M sup f  p  and m inf f  p . Since f  A is closed, by Theorem 3.3.26, M + f  A and m + f  A. Hence, there exists u and ) in A such that f u  M and f ) m  i.e., f u  sup f  p  and f ) inf f  p .
p+ A p+ A p+ A p+ A

Theorem 5.2.25 Suppose that f is a continuous one-to-one mapping of a compact metric space X onto a metric space Y . Then the inverse mapping f 1 which is dened by f 1  f  x  x for all x + X is a continuous mapping that is a one-toone correspondence from Y to X. Proof. Suppose that f is a continuous one-to-one mapping of a compact metric space X onto a metric space Y . Because f is one-to-one, the inverse f 1 is a function from rng  f  Y in X. From the Open Set Characterization of Continuous Functions, we know that f 1 is continuous in Y if f U  is open in Y for every U that is open in X. Suppose that U t X is open. Then, by Theorem 3.3.37, U c is compact as a closed subset of the compact metric space X. In view of Theorem

206

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

5.2.18, f U c  is compact. Since every compact subset of a metric space is closed (Theorem 3.3.35), we conclude that f U c  is closed. Because f is one-to-one, f U c  f  X   f U  then f onto yields that f U c  Y  f U  f U c . Therefore, f U c is closed which is equivalent to f U  being open. Since U was arbitrary, for every U open in X, we have that f U  is open in Y . Hence, f 1 is continuous in Y .

5.2.3 Continuity and Connectedness

Theorem 5.2.26 Suppose that f is a continuous mapping for a metric space X into a metric space Y and E t X. If E is a connected subset of X, then f  E  is connected in Y . Excursion 5.2.27 Fill in what is missing in order to complete the following proof of Theorem 5.2.26. Space for scratch work. Proof. Suppose that f is a continuous mapping from a metric space X into a metric space Y and E t X is such that f  E  is not connected. Then we can let f  E  A C B where A and B are nonempty subsets of Y  i.e., A / 3, B / 3 and
1

ADB H
de f

ADB

3. Consider G

E D f 1  B . Then neither G nor H is empty and b c E D f 1  A C

3

de f

E D f 1  A and

GCH

2 

ED E D f 1  A C B  .
4 b c Because A t A, f 1  A t f 1 A . Since G t f 1  A, the transitivity of containment yields that . From the Corollary to the Open Set

b c Characterization for Continuous Functions, f 1 A is .

6

5

5.2. CONTINUOUS FUNCTIONS ON METRIC SPACES b c b c It follows that G t f 1 A . From G t f 1 A and H t f 1  B , we have that b c G D H t f 1 A D f 1  B  f 1
7

207

f 1
8  9 

The same argument yields that G D H 3. From E G C H , G / 3, H / 3 and G D H G D H 3, we conclude that E is . Hence, for f a continuous mapping from a metric space X into a metric space Y and E t X , if f  E  is not connected, then E is not connected. According to the contrapositive, we conclude that, if , then , as needed.
11 12 10

***Acceptable responses are: (1) separated (2) E D f 1  B , (3) f 1  AC f 1  B , c b (4) E , (5) G t f 1 A , (6) closed, (7) A D B , (8) 3, (9) 3, (10) not connected, (11) E is connected, and (12) f  E  is connected.***

Theorem 5.2.28 Suppose that f is a real-valued function on a metric space  X  d . If f is continuous on S, a nonempty connected subset of X, then the range of f S , denoted by R  f S  , is either an interval or a point. Theorem 5.2.29 (The Intermediate Value Theorem) Let f be a continuous realvalued function on an interval [a  b]. If f a   f b and if c +  f a   f b, then there exist a point x + a  b such that f  x  c. Proof. Let E f [a  b]. Because [a  b] is an interval, from Theorem 3.3.60, we know that [a  b] is connected. By Theorem 5.2.26, E is also connected as the continuous image of a connected set. Since f a  and f b are in E , from Theorem 3.3.60, it follows that if c is a real number satisfying f a   c  f b, then c is in E . Hence, there exists x in [a  b] such that f  x  c. Since f a  is not equal to c and f b is not equal to c, we conclude that x is in a  b. Therefore, there exists x in a  b such that f x  c, as claimed.

208

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

5.3 Uniform Continuity

Our denition of continuity works from continuity at a point. Consequently, point dependency is tied to our =  proofs of limits. For example, if we carried out a 2x 1 =  proof that f  x  is continuous at x 2, corresponding to 0, } x 1 | 1 2x 1 taking = min  will work nicely to show that lim 5 f 2 x 2 x  1 2 6 3 however, it would not work for showing continuity at x . On the other hand, 2 | } 1 corresponding to 0, taking = min  will work nicely to show that 4 24 t u 2x 1 3 lim 8 f . The point dependence of the work is just buried in the 3 x 1 2 x
2

focus on the local behavior. The next concept demands a niceness that is global. Denition 5.3.1 Given metric spaces  X  d X  and Y dY , a function f : X  Y is uniformly continuous on X if and only if e d 1 0 2= 0 1 p  1q   p q + X F d X  p  q   = " dY  f  p   f q    .

Example 5.3.2 The function f x  x 2 : U  U is uniformly continuous on [1 3]. . For x1  x2 + [1 3], the triangular inequality yields For 0 let = 6 that x1 x 2  n x 1  x2  n 6.

Hence, x1  x2 + [1 3] and x 1  x2   = implies that n n n 2 2n  f  x1   f x2  nx1  x2 n x1  x 2  x 1

x2   =  6

Since 0 and x1  x2 + [1 3] were arbitrary, we conclude that f is uniformly continuous on [1 3]. Example 5.3.3 The function f x  U. x 2 : U  U is not uniformly continuous on

5.3. UNIFORM CONTINUITY

209

We want to show that there exists a positive real number such that corresponding to every positive real number = we have (at least two points) x1 x1 = and x2 x2 = for which x 1 =  x2 =  = and  f x1   f  x2  o . This statement is an exact translation of the negation of the denition. For the given function, we want to exploit the fact the as x increases x 2 increase at a rapid (not really uniform) rate. = 1 and Take 1. For any positive real number = , let x 1 x 1 = 2 = 1 x2 x2 = . Then x1 and x 2 are real numbers such that = nt u t un n = 1 1 n n n = = x1  x 2  n  2 = = n 2 while n n n 2 2n  f  x1   f  x2  nx1  x2 n x1  x2  x1 x2  t ut u = = 2 =2 1o1 . 2 2 = 4

Hence, f is not uniformly continuous on U. Example 5.3.4 For p + U1 , let f  p  2 p 1, p 2 . Then f : U1  U2 is uniformly continuous on the closed interval [0 2]. For 0, let = T . If p1 + [0 2] and p2 + [0 2], then 2 5 4 and dU2  f  p1   f  p2   p1 p2 2 4  p1 T 2 p1 p2 2 n 4  p1   p2 2 n 4 b 2 c2 T  2 5 T 20 . 22 20

1  2 p2 T  p1  p2  4  p1

2  p2 p1 2 T p2 2  =  20

12

Since 0 and p1  p2 + [1 2] were arbitrary, we conclude that f is uniformly continuous on [0 2]. Theorem 5.3.5 (Uniform Continuity Theorem) If f is a continuous mapping from a compact metric space X to a metric space Y , then f is uniformly continuous on X.

210

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

Excursion 5.3.6 Fill in what is missing in order to complete the proof of the Uniform Continuity Theorem. Proof. Suppose f is a continuous mapping from a compact metric space  X  d X  to a metric space Y  dY  and that 0 is given. Since f is continuous, for each p + X , there exists a positive real  number = p such that q + X F d X q  p  = p "  dY  f  p  f q   . Let J N  p  : p + X . Since neighborhoods are  1 =p  2 2 open sets, we conclude that J is an . Since X is compact there exists a nite number of elements of J that covers X , say p1  p2   pn . Hence, n > b c Xt N1 pj . =p j j 1 2 j k 1 Let = min = p j . Then, = 0 and the minimum of a nite number of de f 2 1n j nn positive real numbers. Suppose that p q + X are such that d  p q   = . Because p + X and n b c 6 X t N1 p j , there exists a positive integer k , 1 n k n n , such that j 1 =p j 2 1 . Hence, d  p  pk   = pk . From the triangular in2 2  equality d X q  pk  n d X q  p  d X  p  pk   =
3 1

n = pk .

Another application of the triangular inequality and the choices that were made for = p yield that dY  f  p  f q  n
4


5

1 ***Acceptable ll ins are: (1) open cover for X (2) p + N 1  pk , (3) = pk , (4) 2 =k 2 dY  f  p  f  pk  dY  f  pk   f q , . *** 2 2

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS

211

5.4 Discontinuities and Monotonic Functions

Given two metric spaces  X  d X  and Y dY  and a function f from a subset A of X into Y . If p + X and f is not continuous at p, then we can conclude that f is not dened at p ( p +  A dom  f ), lim f  x  does not exist, and/or p + A and
xp
x+A

x p
x+A

lim f x  exists but f  p  /

xp
x+A

lim f  x  a point for which any of the three condi-

tions occurs is called a point of discontinuity. In a general discussion of continuity of given functions, there is no need to discuss behavior at points that are not in the domain of the function consequently, our consideration of points of discontinuity is restricted to behavior at points that are in a specied or implied domain. Furthermore, our discussion will be restricted to points of discontinuity for real-valued functions of a real-variable. This allows us to talk about one-sided limits, behavior on both sides of discontinuities and growth behavior. Denition 5.4.1 A function f is discontinuous at a point c + dom  f  or has a discontinuity at c if and only if either lim f  x  doesnt exist or lim f  x  exists x c x c and is different from f c. Example 5.4.2 The domain of f x  points of discontinuity on its domain. x  is U 
0. Consequently, f has no x

 x  !  , for x + U 
0 x Example 5.4.3 For the function f x  , !  1 , for x 0 dom  f  U and x 0 is a point of discontinuity of f . To see that lim f  x  does not exist, note that, for every positive real number =, n t u n t u n n = = n f  1 and f   f 0n n n 2 2
x 0

2.

1 Hence, if , then, for every positive real number =, there exists x + dom  f  2 such that 0  x   = and  f  x   f 0 o . Therefore, f is not continuous at x 0.

212

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY  t u 1 ! !  x sin x ! !  0 , for x + U 

0 , then g has no dis, for x 0

Example 5.4.4 If g  x  continuities in U.

Excursion 5.4.5 Graph the following function f and nd A and B

x + dom  f  : f is continuous at x 

x + dom  f  : f is discontinuous at x .  1 4x 1 ! ! , xn ! ! 1 x 2 ! ! ! ! 1 ! ! ! 2 , x n1 !  2 f x  2x 4 , 1x n3 ! ! ! ! Jx K 2 , 3x n6 ! ! ! ! ! ! ! !  14  x  10 , 6  x  14 G 14  x  x  14

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS ***Hopefully, your graph revealed that A
3 4 5 6.***

213

U 
1 3 4 5 6 14 and B

Denition 5.4.6 Let f be a function that is dened on the segment a  b. Then, for any point x + [a  b, the right-hand limit is denoted by f x  and s L b c Kr
 f x  q % 1
tn * t t x  b F lim t x " lim f t q     n n n n 1
n * n *

and, for any x + a  b], the left-hand limit is denoted by f  x  and f x  b c Kr *

tn  t a  x  F lim tn q % 1
tn n 1
n *

x " lim f tn 

n *

L q .

Remark 5.4.7 From the treatment of one-sided limits in frosh calculus courses, recall that lim f t  q if and only if
t x

1

0 2=
t x 

= 

d 0 1t  t + dom  f  F x  t  x

= "  f t   q   

and lim f t  1 0 2=

q if and only if =  d e 0 1t  t + dom  f  F x  =  t  x "  f t   q    .

The Sequences Characterization for Limits of Functions justies that these denitions are equivalent to the denitions of f x  and f x , respectively. Excursion 5.4.8 Find f  x  and f  x  for every x + B where B is dened in Excursion 5.4.5.

***For this function, we have f 3 2 f 3  6 f 5 6 f 5  7 and f 6 7 f 6 

5 f 4 7.***

5 f 4 

Remark 5.4.9 For each point x where a function f is continuous, we must have f x  f x  f  x .

214

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

Denition 5.4.10 Suppose the function f is dened on the segment a  b and discontinuous at x + a  b. Then f has a discontinuity of the rst kind at x or a simple discontinuity at x if and only if both f  x  and f  x  exist. Otherwise, the discontinuity is said to be a discontinuity of the second kind. Excursion 5.4.11 Classify the discontinuities of the function f in Excursion 5.4.5.

Remark 5.4.12 The function F  x 

  f x  

, for

x + R 
1 14 where

0 , for x 1 G x 14 f is given in Excursion 5.4.5 has discontinuities of the second kind at x 1 and x 14.

Excursion 5.4.13 Discuss the continuity of each of the following.  2 x x 6 ! !  , x  2 x 2 ! !  2x  1 , x o 2   2 , x rational  1 , x irrational

1. f x 

2. g  x 

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS

215

***Your discussion of (1) combines considers some cases. For *  x  2, x2  x  6 is continuous as the quotient of polynomials for which the denominator x 2 is not going to zero, while continuity of 2x  1 for x 2 follows from the limit of the sum theorem or because 2x  1 is a polynomial consequently, the only point in the domain of f that needs to be checked is x 2. Since f 2  x2  x  6 lim  x  3 5, f 2  lim 2 x  1 5, lim x 2 x 2 x 2 x 2 and f 2 5, it follows that f is also continuous at x 2. That the function given in (2) is not continuous anywhere follows from the density of the rationals and the irrationals each point of discontinuity is a discontinuity of the second kind.*** Denition 5.4.14 Let f be a real-valued function on a segment a  b. Then f is said to be monotonically increasing on a  b if and only if d e 1x1  1x2  x1  x2 + a  b F x1  x2 " f x1  n f  x2  and f is said to be monotonically decreasing on a  b if and only if d e 1x1  1x2  x1  x2 + a  b F x1 x2 " f x1  n f  x2  The class of monotonic functions is the set consisting of both the functions that are increasing and the functions that are decreasing. Excursion 5.4.15 Classify the monotonicity of the function f that was dened in Excursion 5.4.5

***Based on the graph, we have that f is monotonically increasing in each of *u 1, 1 1, and 3 6 the function t is monotonically decreasing in each of  t u 1 1  3 , 6 14, and 14 *. The section  1 is included in both statements 2 2

216

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

because the function is constant there. As an alternative, we could have claimed that t f uis both monotonically increasing and monotonically decreasing in each of 1  1 , 3 4, 4 5, and 5 6 and distinguished the other segments according to 2 the property of being strictly monotonically increasing and strictly monotonically decreasing.*** Now we will show that monotonic functions do not have discontinuities of the second kind. Theorem 5.4.16 Suppose that the real-valued function f is monotonically increasing on a segment a  b. Then, for every x + a  b both f x  and f  x  exist,
a t  x

sup f t 

f  x  n f  x  n f  x 

x t b

inf f t 

and 1x  1 y  a  x  y  b " f  x  n f  y  . Excursion 5.4.17 Fill in what is missing in order to complete the following proof of the theorem. Space for scratch work. Proof. Suppose that f is monotonically increasing on the segment a  b and x + a  b. Then, for every t + a  b such that a  t  x , . Hence, B
de f

f t  : a  t  x  is bounded above by f  x . , the set B has a

2

1 

By the least upper bound let u show that u f x .

sup  B . Now we want to

Let 0 be given. Then u sup  B  and u   u yields the existence of a * + B such that . From the denition of B , * is the image of a point in
4  3

. Let =

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS be such that x  = + a  x  and f  x  = t +  x  = x , then f x  = n f t  and
5

217

* . If .

Since u   * and f x  n u , the transitivity of less than or equal to yields that  f t 

6

and

f t  n
7

Because t was arbitrary, we conclude that 1t   x  =  t  x " u   f t  n u  . Finally, it follows from 1 0
8 t x 

0 being arbitrary that  i.e.,

f x 

lim f t 

u.

For every t + a  b such that x  t  b, we also have that f  x  n f t  from which it follows that
f t  : x  t  b is bounded C by
de f

f x . From the greatest lower bound property of the reals, C has a greatest lower bound that we will denote by ) . Use the space provided to prove that ) f  x .

9

10

218

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

Next suppose that x  y + a  b are such that x  y . Because f  x  lim f t  inf

f t  : x  t  b, x  y  t  x  b and f is monotonically increasing, it follows that inf
f t  : x  t  y  .
11 t x

From our earlier discussion, f  y  lim f t 

t  y

.
12

Now, x  y  t a  y  yields that f  y  Therefore,

13

sup
f t  : x  t  y  . as claimed.

***The expected responses are: (1) f t  n f x , (2) least upper bound property, (3) u d   *  u , (4) a  x , (5) f t  n ef x , (6) u  , (7) u , (8) 2= 0 1t  x  =  t  x " u   f t   u  , (9) below, (10) Let 0 be given. Then ) inf C implies that there exists * + C such that )  *  ) . Since * + C , * is the image of some point in x  b. Let = 0 be such that x = + x  b and f  x = * . Now suppose t +  x  x =. Then f x  n f t  and f t  n f x = * . Since ) n fd x  and *  ) , it follows that ) n f te and f t   ) . Thus, 2= 0 1t  x  t  x = " )  f t  n ) . Because 0 was arbitrary, we conclude that ) lim f t  f  x ., (11) f x , (12) sup
f t  : a  t  y  and (13) f x  n f  y .*** Corollary 5.4.18 Monotonic functions have no discontinuities of the second kind. The nature of discontinuities of functions that are monotonic on segments allows us to identify points of discontinuity with rationals in such a way to give us a limit on the number of them. Theorem 5.4.19 If f is monotonic on the segment a  b, then
x + a  b : f is discontinuous at x  is at most countable.
t x

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS

219

Excursion 5.4.20 Fill in what is missing in order to complete the following proof of Theorem 5.4.19. Proof. Without loss of generality, we assume that f is a function that is monotonically increasing in the segment a  b I . If f is continuous in I , then f has
de f

no points of discontinuity there and we are done. Suppose that f is not continuous on I and let D
* + I : f is not continuous at * . From our assumption D / 3 and we can suppose that ? + D. Then ? + dom  f , 1x   x + I F x  ? " f  x  n f ?  and 1x  x + I F ?  x "
1

From Theorem 5.4.16, f ?  and f ?  exist furthermore, f ?  sup

f x  : x  ?  , f ? 
2

and f ?  n f ? . Since ? is a discontinuity for f , f ? 

3

f ? .

From the Density of the Rationals, it follows that there exists a rational r? such that f ?   r?  f ? . Let Ir?  f ?   f ? . If D 
?  3, then  D 1 and we are done. If D 
?  / 3 then we can choose another G + D such that G / ? . Without loss of generality suppose that G + D is such that ?  G . Since ? was an arbitrary point in the discussion just completed, we know that there exists a rational rG , rG / r? , such that and we can let IrG  f G   f G . Since ?  G , it follows from Theorem 5.4.16 that n f G . Thus, Ir? D IrG . 5  6 j k Now, let J Ir< : < + D and H : D  J be dened by H * Ir* . Now we claim that H is a one-to-one correspondence. To see that H is , suppose that *1  *2 + D and H *1  H *2 .
4

220 Then

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

7 

To see that H is onto, note that by denition H  D  t J and suppose that A + J . Then
8 

Finally, H : D  J yields that D q J . Since r< + T for each Ir< + J , we have that J  n T 80 . Therefore,  D  n 80  i.e., D is .
9

11

***Acceptable responses are : (1) f ?  n f  x , (2) inf

f x  : ?  x , (3) , (4) f G   rG  f G , (5) f ? , (6) 3, (7) Ir*1 Ir*2 . From the Trichotomy Law, we know that one and only one of *1  *2 , *1 *2 , or *2  *1 holds. Since either *1  *2 or *2  *1 implies that Ir*1 D Ir*2 3, we conclude that *1 *2 . Since *1 and *2 were arbitrary, 1*1  1*2  [ H *1  H *2  " *1 *2 ] i.e., H is one-to-one., (8) there exists r + T such that A Ir and r +  f D  f D  for some D + D . It follows that H D A or A + H  D . Since A was an arbitrary element of J , we have that 1 A [ A + J " A + H  D ] i.e., J t H  D . From H  D  t J and J t H  D , we conclude that J H  D . Hence, H is onto., and (9) at most countable.*** Remark 5.4.21 The level of detail given in Excursion 5.4.20 was more that was needed in order to offer a well presented argument. Upon establishing the ability to associate an interval Ir? with each ? + D that is labelled with a rational and justifying that the set of such intervals is pairwise disjoint, you can simply assert that you have established a one-to-one correspondence with a subset of the rationals and the set of discontinuities from which it follows that the set of discontinuities is at most countable. I chose the higher level of detailwhich is also acceptablein

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS

221

order to make it clearer where material prerequisite for this course is a part of the foundation on which we are building. For a really concise presentation of a proof of Theorem 5.4.19, see pages 96-97 of our text. Remark 5.4.22 On page 97 of our text, it is noted by the author that the discontinuities of a monotonic function need not be isolated. In fact, given any countable subset E of a segment a  b, he constructs a function f that is monotonic on a  b with E as set of all discontinuities of f in a  b. More consideration of the example is requested in our exercises.

5.4.1

Limits of Functions in the Extended Real Number System

Recall the various forms of denitions for limits of real valued functions in relationship to innity: Suppose that f is a real valued function on U, c is a real number, and L real number, then 
x *

lim f x 

L % 1

0 2 K % 1

0  x 0 2K

K "  f x   L    0 x K " f x  + N  L 

x *

lim f x 

L % 1

0 2 K % 1

0  x   K "  f  x   L    0 x   K " f x  + N  L 

0 2K

 lim f x 
x c

* % 1 M + U 2=

0 0  x  c  = " f x  M  b c d % 1 M + U 2= 0 x + N= c " f x  M 0 0  x  c  = " f x   M  b c d % 1 M + U 2= 0 x + N= c " f x   M

d where N= c denotes the deleted neighborhood of c, N= c 

c.

 lim f x 
x c

* % 1 M + U 2=

Based of the four that are given, complete each of the following. 
x *

lim f x 

*%

222 

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

x *

lim f x 

* %

x *

lim f x 

*%

x *

lim f x 

* %

Hopefully, the neighborhood formulation and the pattern of the various statements suggests that we could pull things together if we had comparable descriptions for neighborhoods of * and *. Denition 5.4.23 For any positive real number K , N K *
x + UC
* * : x K * and

x + UC
* * : x  K  are neighborhoods of and N K * *, respectively. With this notation we can consolidate the above denitions.

Denition 5.4.24 Let f be a real valued function dened on U. Then for A and c in the extended real number system, lim f x  A if and only if for every neighborhood of A, N  A there exists a deleted neighborhood of c, N `d c, such that x + N `d c implies that f  x  + N  A. When specication is needed this will be referred to a the limit of a function in the extended real number system. Hopefully, the motivation that led us to this denition is enough to justify the claim that this denition agrees with the denition of lim f  x  A when c and A x c are real. Because the denition is the natural generalization and our proofs for the properties of limits of function built on information concerning neighborhoods, we note that we can establish some of the results with only minor modication in the proofs that have gone before. We will simply state analogs.
x c

5.4. DISCONTINUITIES AND MONOTONIC FUNCTIONS

223

Theorem 5.4.25 Let f be a real-valued function that is dened on a set E t U and suppose that lim f t  A and lim f t  C for c, A, and C in the extended t c t c real number system. Then A C. Theorem 5.4.26 Let f and g be real-valued functions that are dened on a set E t U and suppose that lim f t  A and lim g t  B for c, A, and B in the t c t c extended real number system. Then 1. lim  f
t c t c

g  t 

B,

2. lim  f g  t  t u f 3. lim t  t c g

AB, and A B

whenever the right hand side of the equation is dened. Remark 5.4.27 Theorem 5.4.26 is not applicable when the algebraic manipula* A tions lead to the need to consider any of the expressions *  *, 0  *, , or * 0 because none of these symbols are dened. The theorems in this section have no impact on the process that you use in order to nd limits of real functions as x goes to innity. At this point in the coverage of material, given a specic function, we nd the limit as x goes to innity by using simple algebraic manipulations that allow us to apply our theorems for algebraic combinations of functions having nite limits. We close this chapter with two examples that are intended as memory refreshers. c b c b 2 i x 3 x sin x x  3x 3 5 Example 5.4.28 Find lim . x * 4x 3  7 Since the given function is the quotient of two functions that go to innity as x goes to innity, we factor in order to transform the given in to the quotient of functions that will have nite limits. In particular, we want to make use of the fact 1 that, for any p + J , lim p 0. From x * x t u t u 1 5 sin x b 2 c b c 3 i 1 x  3x 3 5 i x 3 x sin x x x3 x2 t u lim lim , x * x * 7 4x 3  7 4 3 x

224

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

The limit of the quotient and limit of the sum theorem yields that b 2 c b c x  3x 3 5 i x 3 x sin x 0  3 0 i 1 0 lim x * 4x 3  7 40 rT s T Example 5.4.29 Find lim 2x 2 x 2  2x 2  x  1 .
x *

3 i . 4

In its current form, it looks like the function is tending to *  * which is undened. In this case, we will try unrationalizing the expression in order to get a quotient at will allow some elementary algebraic manipulations. Note that s rT T 2 2 2x x 2  2x  x  1 rT s rT s T T 2x 2 x 2  2x 2  x  1 2x 2 x 2 2x 2  x  1 rT s T 2x 2 x 2 2x 2  x  1 c b c b 2 x 2  2x 2  x  1 2x rT s T 2x 2 x 2 2x 2  x  1 2x 3 rT s. T 2 2 2x x 2 2x  x  1 T Furthermore, for x  0, x 2 x  x. Hence, rT s T 2 2 lim 2x x 2  2x  x  1 x * 2x 3 s lim rT T x * 2x 2 x 2 2x 2  x  1 2x 3 U lim U x * T 1 2 1 1 2 x 2 2   x x2 x x2 3 2 x lim 1 U U x * 1 2 1 1 2 2   x x2 x x2 T t u 2 1  2 T T . 1 T 2 2 2 2

5.5. PROBLEM SET E

225

5.5 Problem Set E

1. For each of the following real-valued functions of a real variable give a wellwritten =  proof of the claim. b (a) lim 3x 2  2x
x 2

(b) lim 8 x 2
x 1 x 16

8 4

(c) lim (d) lim

3 3 x 1 x  2 x 4 (e) lim 7. x 3 2 x  5

2. For each of the following real-valued functions of a real variable nd the implicit domain and range. (a) f  x  (b) f x  (c) f  x  sin x x2  1 T 2x 1 x 2 x 5x 6  3 ! ! ! ! x 3 ! ! !  x  2 ! ! x 2 ! ! ! ! !  1 , x 0 2

3. Let f  x 

, 0n x 2Fx , x 2

(a) Sketch a graph for f  (b) Determine where the function f is continuous.

226

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY n n n  n 2 7  nx  5 x  6n , for n x  2 n o  T 36  6x

x 2 1 x 1 5 2

4. Let f x    

n n n , for n x  7 2  1 . 1 1.

and
5 2

, for , for

g x 

x/ x

(a) Discuss the continuity of f at x (b) Discuss the continuity of  f g  x  5. For f : F  U given by f z  T 2. 6. When it exists, nd t (a) lim
x 2

f x g  x  at x

1.
z  1 i 

z  give a =  proof that

lim

f z 

x2  4 T 2  3x x 2

u 2 x  1 5 x 1
x a

(b) lim

x 1

T x 1  x4 x 2 3x  4

7. Let f : U  U and suppose that lim f  x 

x a

L L

0. Prove that

lim

f x 

8. Using only appropriate denitions and elementary bounding processes, prove that if g is a real-valued function on U such that lim g  x  M / 0, then x a 1 1 lim . x a [g  x ]2 M2 9. Suppose that A is a subset of a metric space  S  d , f : A  U1 , and g : A  U1 . Prove each of the following. (a) If c is a real number and f  p  c for all p + A, then, for any limit point p0 of A, we have that lim f  p c.
p  p0
p+ A

5.5. PROBLEM SET E

227

(b) If f  p  g  p for all p + A 

p0  were p0 is a limit point of A and lim f  p L , then lim g  p L .
p  p0
p+ A

p  p0
p+ A

(c) If f  p  n g  p  for all p + A, lim f  p 

p  p0
p+ A

L and lim g  p
p  p0
p+ A

M,

then L n M .

10. For each of the following functions on U2 , determine whether or not the given function is continuous at 0 0. Use =  proofs to justify continuity or show lack of continuity by justifying that the needed limit does not exist.  x y2 ! ! ! , for x  y  / 0 0 b c  2 2 2 x y (a) f x  y  ! ! !  0 , for x y 0  ! x 3 y3 ! ! , for x  y  / 0 0 c  b 2 2 2 x y (b) f x  y  ! ! !  0 , for x y 0  ! x 2 y4 ! ! , for x  y  / 0 0 c  b 2 4 2 x y (c) f x  y  ! ! !  0 , for x y 0 11. Discuss the uniform continuity of each of the following on the indicated set. (a) f  x  (b) f  x  x2 1 in the interval [4 9]. 2x 3 x 3 in [1 *.

12. For a  b, let F [a  b] denote the set of all real valued functions that are continuous on the interval [a  b]. Prove that d  f  g  max  f x   g x  is a metric on F [a  b].
a n x nb

13. Correctly formulate the monotonically decreasing analog for Theorem 5.4.16 and prove it.

228

CHAPTER 5. FUNCTIONS ON METRIC SPACES AND CONTINUITY

14. Suppose d that f is monotonically increasing on e a segment I a  b and that 2 M  M + U F 1x   x + I " f x  n M  . Prove that there exists a real number C such that C n M and f b C . 15. A function f dened on an interval I [a  b] is called strictly increasing f x 2  whenever x1 x2 for x1  x2 + I . Furon I if and only if f  x1  thermore, a function f is said to have the intermediate value properly in I if and only if for each c between f a  and f b there is an x0 + I such that f x0  c. Prove that a function f that is strictly increasing and has the intermediate value property on an interval I [a  b] is continuous on a  b. 16. Give an example of a real-valued function f that is continuous and bounded on [0 * while not satisfying the Extreme Value Theorem. 17. Suppose that f is uniformly continuous on the intervals I1 and I2 . Prove that f is uniformly continuous on S I1 C I2 . 18. Suppose that a real-valued function f is continuous on I i where I If f a  and f b exist, show that the function  ! f a  , for x a ! !  f x  , for a  x  b f 0 x  ! ! !  f b , for x b is uniformly continuous on I . 19. If a real valued function f is uniformly continuous on the half open interval 0 1], is it true that f is bounded there. Carefully justify the position taken. [a  b].