4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 85

2.3 The Precise Definition of a Limit 85

2.3 The Precise Definition of a Limit

Now that we have gained some insight into the limit concept, working intuitively with the
informal definition, we turn our attention to its precise definition. We replace vague
phrases like “gets arbitrarily close to” in the informal definition with specific conditions
that can be applied to any particular example. With a precise definition we will be able to
prove conclusively the limit properties given in the preceding section, and we can establish
other particular limits important to the study of calculus.
To show that the limit of ƒ(x) as x : x0 equals the number L, we need to show that the gap
between ƒ(x) and L can be made “as small as we choose” if x is kept “close enough” to x0 .
Let us see what this would require if we specified the size of the gap between ƒ(x) and L.

EXAMPLE 1 A Linear Function

y Consider the function y = 2x - 1 near x0 = 4. Intuitively it is clear that y is close to 7
y  2x  1 when x is close to 4, so limx:4 s2x - 1d = 7. However, how close to x0 = 4 does x have
to be so that y = 2x - 1 differs from 7 by, say, less than 2 units?
Upper bound:
y9 Solution We are asked: For what values of x is ƒ y - 7 ƒ 6 2? To find the answer we
9 first express ƒ y - 7 ƒ in terms of x:
To satisfy 
this 7
 ƒ y - 7 ƒ = ƒ s2x - 1d - 7 ƒ = ƒ 2x - 8 ƒ .
5
Lower bound: The question then becomes: what values of x satisfy the inequality ƒ 2x - 8 ƒ 6 2? To
y5
find out, we solve the inequality:
x ƒ 2x - 8 ƒ 6 2
0 3 4 5
2x - 8 6 2




-2 6
Restrict
to this 6 6 2x 6 10
3 6 x 6 5
FIGURE 2.12 Keeping x within 1 unit
of x0 = 4 will keep y within 2 units of
-1 6 x - 4 6 1.
y0 = 7 (Example 1). Keeping x within 1 unit of x0 = 4 will keep y within 2 units of y0 = 7 (Figure 2.12).

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 86

86 Chapter 2: Limits and Continuity

y In the previous example we determined how close x must be to a particular value x0 to

ensure that the outputs ƒ(x) of some function lie within a prescribed interval about a limit
1
value L. To show that the limit of ƒ(x) as x : x0 actually equals L, we must be able to show
L that the gap between ƒ(x) and L can be made less than any prescribed error, no matter how
10
f(x)
f(x) lies small, by holding x close enough to x0 .
L in here

L
1 Definition of Limit
10
for all x  x0 Suppose we are watching the values of a function ƒ(x) as x approaches x0 (without taking on
in here the value of x0 itself). Certainly we want to be able to say that ƒ(x) stays within one-tenth of
  a unit of L as soon as x stays within some distance d of x0 (Figure 2.13). But that in itself is
x
x not enough, because as x continues on its course toward x0 , what is to prevent ƒ(x) from jit-
0 x0   x0 x0  
tering about within the interval from L - (1>10) to L + (1>10) without tending toward L?
We can be told that the error can be no more than 1>100 or 1>1000 or 1>100,000.
FIGURE 2.13 How should we define
Each time, we find a new d-interval about x0 so that keeping x within that interval satisfies
d 7 0 so that keeping x within the interval
the new error tolerance. And each time the possibility exists that ƒ(x) jitters away from L at
sx0 - d, x0 + dd will keep ƒ(x)
some stage.
within the interval aL - b?
1 1
10
,L +
10
The figures on the next page illustrate the problem. You can think of this as a quarrel
between a skeptic and a scholar. The skeptic presents P-challenges to prove that the limit
does not exist or, more precisely, that there is room for doubt, and the scholar answers
every challenge with a d-interval around x0 .
y How do we stop this seemingly endless series of challenges and responses? By prov-
ing that for every error tolerance P that the challenger can produce, we can find, calculate,
or conjure a matching distance d that keeps x “close enough” to x0 to keep ƒ(x) within that
L
tolerance of L (Figure 2.14). This leads us to the precise definition of a limit.

f(x) lies
L in here
f(x) DEFINITION Limit of a Function
L Let ƒ(x) be defined on an open interval about x0 , except possibly at x0 itself. We
say that the limit of ƒ(x) as x approaches x0 is the number L, and write
lim ƒsxd = L,
for all x  x 0 x:x0
in here
if, for every number P 7 0, there exists a corresponding number d 7 0 such that
 
for all x,
x
x
0
x0   x0 x0   0 6 ƒ x - x0 ƒ 6 d Q ƒ ƒsxd - L ƒ 6 P.

FIGURE 2.14 The relation of d and P in

the definition of limit. One way to think about the definition is to suppose we are machining a generator
shaft to a close tolerance. We may try for diameter L, but since nothing is perfect, we must
be satisfied with a diameter ƒ(x) somewhere between L - P and L + P. The d is the
measure of how accurate our control setting for x must be to guarantee this degree of accu-
racy in the diameter of the shaft. Notice that as the tolerance for error becomes stricter, we
may have to adjust d. That is, the value of d, how tight our control setting must be, de-
pends on the value of P, the error tolerance.

Examples: Testing the Definition

The formal definition of limit does not tell how to find the limit of a function, but it en-
ables us to verify that a suspected limit is correct. The following examples show how the

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 87

2.3 The Precise Definition of a Limit 87

y y y y

y  f (x) y  f (x) y  f (x) y  f(x)

1 1
L L
10 10 1 1
L L
100 100
L L L L
1 1
1 1 L L
L 100 100
10 L
10

x x x x
0 x0 0 x0 0 x0 0 x0
x 0  1/10 x 0  1/10 x 0  1/100 x 0  1/100
The challenge: Response: New challenge: Response:
Make f (x) – L    1 x  x 0  1/10 (a number) Make f (x) – L    1 x  x 0  1/100
10 100

y y
y  f (x) y  f (x)
1 1
L L
1000 1000

L L
1 1
L L
1000 1000

x x
0 x0 0 x0

New challenge: Response:

 1 x  x 0  1/1000
1000

y y y
y  f(x) y  f (x) y  f (x)
1 1
L L
100,000 100,000
L L L
1 1
L L
100,000 100,000

x x x
0 x0 0 x0 x0
0
New challenge: Response: New challenge:
1

100,000 x  x 0  1/100,000   ...

definition can be used to verify limit statements for specific functions. (The first two ex-
amples correspond to parts of Examples 7 and 8 in Section 2.1.) However, the real purpose
of the definition is not to do calculations like this, but rather to prove general theorems so
that the calculation of specific limits can be simplified.

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 88

88 Chapter 2: Limits and Continuity

y EXAMPLE 2 Testing the Definition

y  5x  3
Show that
2
lim s5x - 3d = 2.
x:1
2
Solution Set x0 = 1, ƒsxd = 5x - 3, and L = 2 in the definition of limit. For any given
2 P 7 0, we have to find a suitable d 7 0 so that if x Z 1 and x is within distance d of
x0 = 1, that is, whenever
x
0 1 1 1 0 6 ƒ x - 1 ƒ 6 d,
5 5
it is true that ƒ(x) is within distance P of L = 2, so

ƒ ƒsxd - 2 ƒ 6 P.
We find d by working backward from the P-inequality:
–3
NOT TO SCALE ƒ s5x - 3d - 2 ƒ = ƒ 5x - 5 ƒ 6 P
5ƒx - 1ƒ 6 P
FIGURE 2.15 If ƒsxd = 5x - 3 , then
0 6 ƒ x - 1 ƒ 6 P>5 guarantees that
ƒ x - 1 ƒ 6 P>5.
ƒ ƒsxd - 2 ƒ 6 P (Example 2). Thus, we can take d = P>5 (Figure 2.15). If 0 6 ƒ x - 1 ƒ 6 d = P>5, then

ƒ s5x - 3d - 2 ƒ = ƒ 5x - 5 ƒ = 5 ƒ x - 1 ƒ 6 5sP>5d = P,
which proves that limx:1s5x - 3d = 2.
The value of d = P>5 is not the only value that will make 0 6 ƒ x - 1 ƒ 6 d imply
ƒ 5x - 5 ƒ 6 P. Any smaller positive d will do as well. The definition does not ask for a
“best” positive d, just one that will work.

y EXAMPLE 3 Limits of the Identity and Constant Functions

yx Prove:
x0  
(a) lim x = x0 (b) lim k = k (k constant).
x0   x:x0 x:x0
x0
x0   Solution

x0  
(a) Let P 7 0 be given. We must find d 7 0 such that for all x
0 6 ƒ x - x0 ƒ 6 d implies ƒ x - x0 ƒ 6 P.
x The implication will hold if d equals P or any smaller positive number (Figure 2.16).
0 x0   x0 x0  
This proves that limx:x0 x = x0 .
FIGURE 2.16 For the function ƒsxd = x , (b) Let P 7 0 be given. We must find d 7 0 such that for all x
we find that 0 6 ƒ x - x0 ƒ 6 d will
0 6 ƒ x - x0 ƒ 6 d implies ƒ k - k ƒ 6 P.
guarantee ƒ ƒsxd - x0 ƒ 6 P whenever
d … P (Example 3a). Since k - k = 0, we can use any positive number for d and the implication will hold
(Figure 2.17). This proves that limx:x0 k = k.

Finding Deltas Algebraically for Given Epsilons

In Examples 2 and 3, the interval of values about x0 for which ƒ ƒsxd - L ƒ was less than
P was symmetric about x0 and we could take d to be half the length of that interval.

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 89

2.3 The Precise Definition of a Limit 89

y When such symmetry is absent, as it usually is, we can take d to be the distance from x0
to the interval’s nearer endpoint.
yk
k
k EXAMPLE 4 Finding Delta Algebraically
k
For the limit limx:5 2x - 1 = 2, find a d 7 0 that works for P = 1. That is, find a
d 7 0 such that for all x

x
0 6 ƒx - 5ƒ 6 d Q ƒ 2x - 1 - 2 ƒ 6 1.
0 x0   x0 x0  
Solution We organize the search into two steps, as discussed below.
FIGURE 2.17 For the function ƒsxd = k ,
1. Solve the inequality ƒ 2x - 1 - 2 ƒ 6 1 to find an interval containing x0 = 5 on
we find that ƒ ƒsxd - k ƒ 6 P for any
which the inequality holds for all x Z x0 .
positive d (Example 3b).
ƒ 2x - 1 - 2 ƒ 6 1
-1 6 2x - 1 - 2 6 1
1 6 2x - 1 6 3
1 6 x - 1 6 9
2 6 x 6 10

3 3
The inequality holds for all x in the open interval (2, 10), so it holds for all x Z 5 in
x this interval as well (see Figure 2.19).
2 5 8 10
2. Find a value of d 7 0 to place the centered interval 5 - d 6 x 6 5 + d (centered
FIGURE 2.18 An open interval of at x0 = 5) inside the interval (2, 10). The distance from 5 to the nearer endpoint of
radius 3 about x0 = 5 will lie inside the (2, 10) is 3 (Figure 2.18). If we take d = 3 or any smaller positive number, then the
open interval (2, 10). inequality 0 6 ƒ x - 5 ƒ 6 d will automatically place x between 2 and 10 to make
ƒ 2x - 1 - 2 ƒ 6 1 (Figure 2.19)

y
0 6 ƒx - 5ƒ 6 3 Q ƒ 2x - 1 - 2 ƒ 6 1.

y  兹x  1
3 How to Find Algebraically a D for a Given f, L, x0 , and P>0
The process of finding a d 7 0 such that for all x
2 0 6 ƒ x - x0 ƒ 6 d Q ƒ ƒsxd - L ƒ 6 P
can be accomplished in two steps.
1
1. Solve the inequality ƒ ƒsxd - L ƒ 6 P to find an open interval (a, b) contain-
3 3 ing x0 on which the inequality holds for all x Z x0 .
x
0 1 2 5 8 10 2. Find a value of d 7 0 that places the open interval sx0 - d, x0 + dd centered
NOT TO SCALE at x0 inside the interval (a, b). The inequality ƒ ƒsxd - L ƒ 6 P will hold for all
x Z x0 in this d-interval.
FIGURE 2.19 The function and intervals
in Example 4.

EXAMPLE 5 Finding Delta Algebraically

Prove that limx:2 ƒsxd = 4 if
x 2,
ƒsxd = e
x Z 2
1, x = 2.

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 90

90 Chapter 2: Limits and Continuity

y
Solution Our task is to show that given P 7 0 there exists a d 7 0 such that for all x
y  x2

4
0 6 ƒx - 2ƒ 6 d Q ƒ ƒsxd - 4 ƒ 6 P.
1. Solve the inequality ƒ ƒsxd - 4 ƒ 6 P to find an open interval containing x0 = 2 on
4 (2, 4) which the inequality holds for all x Z x0 .
For x Z x0 = 2, we have ƒsxd = x 2 , and the inequality to solve is ƒ x 2 - 4 ƒ 6 P:
4
ƒ x2 - 4 ƒ 6 P
(2, 1)
-P 6 x 2 - 4 6 P
4 - P 6 x2 6 4 + P
x
0 2 24 - P 6 ƒ x ƒ 6 24 + P Assumes P 6 4 ; see below.
兹4   兹4  
24 - P 6 x 6 24 + P. An open interval about x0 = 2
FIGURE 2.20 An interval containing that solves the inequality

The inequality ƒ ƒsxd - 4 ƒ 6 P holds for all x Z 2 in the open interval A 24 - P,

x = 2 so that the function in Example 5

24 + P B (Figure 2.20).
satisfies ƒ ƒsxd - 4 ƒ 6 P .

2. Find a value of d 7 0 that places the centered interval s2 - d, 2 + dd inside the in-
terval A 24 - P, 24 + P B .
Take d to be the distance from x0 = 2 to the nearer endpoint of A 24 - P, 24 + P B .
In other words, take d = min E 2 - 24 - P, 24 + P - 2 F , the minimum (the smaller)
of the two numbers 2 - 24 - P and 24 + P - 2. If d has this or any smaller positive
value, the inequality 0 6 ƒ x - 2 ƒ 6 d will automatically place x between 24 - P and
24 + P to make ƒ ƒsxd - 4 ƒ 6 P. For all x,

0 6 ƒx - 2ƒ 6 d Q ƒ ƒsxd - 4 ƒ 6 P.
This completes the proof.
Why was it all right to assume P 6 4? Because, in finding a d such that for all
x, 0 6 ƒ x - 2 ƒ 6 d implied ƒ ƒsxd - 4 ƒ 6 P 6 4, we found a d that would work for
any larger P as well.
Finally, notice the freedom we gained in letting d = min {2 - 24 - P,
24 + P - 2}. We did not have to spend time deciding which, if either, number was the
smaller of the two. We just let d represent the smaller and went on to finish the argument.

Using the Definition to Prove Theorems

We do not usually rely on the formal definition of limit to verify specific limits such as
those in the preceding examples. Rather we appeal to general theorems about limits, in
particular the theorems of Section 2.2. The definition is used to prove these theorems
(Appendix 2). As an example, we prove part 1 of Theorem 1, the Sum Rule.

EXAMPLE 6 Proving the Rule for the Limit of a Sum

Given that limx:c ƒsxd = L and limx:c gsxd = M, prove that

lim sƒsxd + gsxdd = L + M.

x:c

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4709 AWL/ThomasET_ch02p067-143 1/17/05 8:13 AM Page 91

2.3 The Precise Definition of a Limit 91

Solution Let P 7 0 be given. We want to find a positive number d such that for all x
0 6 ƒx - cƒ 6 d Q ƒ ƒsxd + gsxd - sL + Md ƒ 6 P.
Regrouping terms, we get

ƒ ƒsxd + gsxd - sL + Md ƒ = ƒ sƒsxd - Ld + sgsxd - Md ƒ

Triangle Inequality:
… ƒ ƒsxd - L ƒ + ƒ gsxd - M ƒ .
ƒa + bƒ … ƒaƒ + ƒbƒ

Since limx:c ƒsxd = L, there exists a number d1 7 0 such that for all x
0 6 ƒ x - c ƒ 6 d1 Q ƒ ƒsxd - L ƒ 6 P>2.
Similarly, since limx:c gsxd = M, there exists a number d2 7 0 such that for all x
0 6 ƒ x - c ƒ 6 d2 Q ƒ gsxd - M ƒ 6 P>2.
Let d = min 5d1, d26, the smaller of d1 and d2 . If 0 6 ƒ x - c ƒ 6 d then ƒ x - c ƒ 6 d1 ,
so ƒ ƒsxd - L ƒ 6 P>2, and ƒ x - c ƒ 6 d2 , so ƒ gsxd - M ƒ 6 P>2. Therefore

P P
ƒ ƒsxd + gsxd - sL + Md ƒ 6 2 + 2 = P.

This shows that limx:c sƒsxd + gsxdd = L + M.

Let’s also prove Theorem 5 of Section 2.2.

EXAMPLE 7
Given that limx:c ƒsxd = L and limx:c gsxd = M, and that ƒsxd … gsxd for all x in an
open interval containing c (except possibly c itself), prove that L … M.

Solution We use the method of proof by contradiction. Suppose, on the contrary, that
L 7 M. Then by the limit of a difference property in Theorem 1,

lim s gsxd - ƒsxdd = M - L.

x:c
Therefore, for any P 7 0, there exists d 7 0 such that

ƒ sgsxd - ƒsxdd - sM - Ld ƒ 6 P whenever 0 6 ƒ x - c ƒ 6 d.

Since L - M 7 0 by hypothesis, we take P = L - M in particular and we have a number

d 7 0 such that

ƒ sg sxd - ƒsxdd - sM - Ld ƒ 6 L - M whenever 0 6 ƒ x - c ƒ 6 d.

Since a … ƒ a ƒ for any number a, we have

sgsxd - ƒsxdd - sM - Ld 6 L - M whenever 0 6 ƒx - cƒ 6 d

which simplifies to

gsxd 6 ƒsxd whenever 0 6 ƒ x - c ƒ 6 d.

But this contradicts ƒsxd … gsxd. Thus the inequality L 7 M must be false. Therefore
L … M.

Copyright © 2006 Pearson Education, Inc., publishing as Pearson Addison-Wesley