6

Limit of functions
6.1 Intervals and Neighborhoods
Open interval (a, b) contains all points x such that a < x < b. Closed interval [a, b]
contains all points x such that a ≤ x ≤ b. A neighborhood of x0 includes all point x in
an open interval containing x0 . Symmetric interval |x − x0 | < ,  > 0, contains all points
x such that x0 −  < x < x0 + .
Interval a < x < b is open ended in the sense that, in theory, number x can be selected
willfully close to the end points a and b yet no x values exist that are the least and the most
in the interval. The set of numbers a < x < b is obviously bounded below and above. Its
least upper bound (lub) or supremum is b; sup(a < x < b) = b. Its greatest lower
bound (glb) or infimum is a; inf(a < x < b) = a. The interval a < x < b does not
contain a minimal value nor a maximal value since both its supremum and infimum are not
in the interval. On the other hand the set of numbers a ≤ x ≤ b attains the minimum a;
min(a ≤ x ≤ b) = a and the maximum b; max(a ≤ x ≤ b) = b.
Real number α is an upper bound on function f (x), a ≤ x ≤ b if α ≥ f (x), a ≤ x ≤ b.
Number β is a lower bound on function f (x), a ≤ x ≤ b if β ≤ f (x), a ≤ x ≤ b.
Theorem: A function bounded above has a least upper bound, sup f . A function bounded
below has a greatest lower bound, inf f .
Proof. Suppose we can, at least theoretically, resolve with α
certainty the question if a given number is un upper bound
on function f (x) or not. suppose that two numbers α and β, sup f
(α+β)/2
α > β are available such that α is an upper bound on f (x)
but β is not. See the figure below. If number γ = (α + β)/2
β
is an upper bound on f (x), then α is replaced by γ < α
and bisection is repeated. Otherwise, if γ is not an upper f(x)
bound on f (x), then β is replaced by γ > β and bisection
is repeated. The limit of this bisection process is sup f since any lesser number is not an
upper bound, and any greater number is an upper bound. In the same way we get inf f .End
of proof.
The last theorem is a paraphrase of the property of real numbers stating that a bounded
non empty set of, finite or infinite, real numbers has a least upper bound and a greatest
lower bound. The empty set is excluded to avoid dealing with upper and lower bounds on
nothing (no-thing).
Remark. Notice that the previous theorem does not claim that if u = sup f (x) on the
interval I = [a, b], then c exists on I such that f (c) = u.
6.2 Limits of Functions
Seeking the limit of function f (x) at point a we make the underlying assumption that
function f (x) is defined for all x = a in some neighborhood of a, but not necessarily at point
a itself. We forbear requiring the less stringent condition that point a is such that in every

1
 Chapter 6
neighborhood of point a, excluding a itself, there is at least one x for which f (x) is defined.
Such point a is called an accumulation point of the interval for the function.
Definition of the limit according to Cauchy: Let f (x) be defined for any x in some
neighborhood of a, except possibly at a itself. The statement limx→a f (x) = l means that
f (x) can be restricted to a willfully small neighborhood of l with x restricted a sufficiently
small neighborhood of a.
Figuratively speaking the limit of function f (x) as x → a is the value of f (x) immediately
to the left of a and equally the value of f (x) immediately to the right of a. Sometimes these
are written as f (a − 0) and f (a + 0), respectively. We understand immediately to mean
the next moment in time without delay. But quantitatively immediately is paradoxical, for
if the temporal extent between the two moments is zero, then they are the same, and if
there is a stretch of time between them, then there is no immediacy. It has all to do with
the basic arithmetical fact that numbers exist, that while being different from zero, may be
arbitrarily close to it. In calculus we are in constant need to move but in a movement that
is imperceptibly small. Calculus assumes that the very small is the ultimately small, and
that the very large is the ultimately large, so as to remove the uncertainty of scale.

y y
y1=l+ε
y1=l+ε
l l
y=f(x) y0=l-ε
y0=l-ε
x x
x0 x a x1 x0 a x1

f(x) x

l-ε l l+ε a

l-ε l l+ε a

l-ε l l+ε a

l-ε l+ε
l a

Say we wish to compute the value of y = x2 immediately to the right and left of x = 2. To
do this we undertake the computation of y = f (x) at x = 2+. We are now confronted by the
apparently contradicting demands that  be different from zero yet be practically zero. We
may not select any definite numerical value for  lest it may not be small enough for x = 2+
2
Chapter 6

to qualify for being immediate to 2. So instead we compute y(2 + ) = (2 + )2 = 4 + 4 + 2

and reason that that as  → 0, f (2 + ) → 4. The values of f (x) = x2 immediately to the
right and to the left of x = 2 are y = 4. It also happens that y(2) is defined at x = 2 and
y(x = 2) = 4. But, this is not always so and the limit of f (x) as x → a may not equal f (a).
We readily verify, for example, that for the function

 2
x , x = 0
f (x) =


1, x=0

limx→0 f (x) = 0, whereas f (0) = 1. Yet, for this same function limx→a f (x) = f (a) = a2 , if
a = 0.
Examples and counterexamples.
1. We will verify that limδx→0 (4 + 3δx) = 4. Say we wish to limit f (δx) = 4 + 3δx to the
interval 4 −  ≤ f (δx) ≤ 4 + , where  > 0. Then

4 −  ≤ 4 + 3δx ≤ 4 + 

and it readily results that −/3 ≤ δx ≤ /3.

2. We shall verify that limδx→0 1/(2 − 3δx) = 1/2. Say we wish to limit f (δx) = 1/(2 − 3δx)
to the interval 1/2 −  ≤ f (δx) ≤ 1/2 + , where  > 0. Then

1 1 1
−≤ ≤ +
2 2 − 3δx 2
or
1 1
− ≤ − ≤
2 − 3δx 2
which by common denominator becomes

3δx
− ≤ ≤ .
4 − 6δx
Restricting δx to be such that 4 − 6δx > 0 we obtain, through multiplication by 4 − 6δx,
that
4 4
−  ≤ δx ≤ 
3 − 6 3 + 6
provided that  < 1/2.
√ √
3. We shall verify that limδx→0 4 − 3δx = 2. Say we wish to limit f (δx) = 4 − 3δx to
the interval 2 −  ≤ f (δx) ≤ 2 + , where  > 0. For this, δx need be such that
√
2 −  ≤ 4 − 3δx ≤ 2 + .

If  < 2, then we may square the inequalities so as to have

4 − 4 + 2 ≤ 4 − 3δx ≤ 4 + 4 + 2
3
 Chapter 6
or
1 1
(−4 − ) ≤ δx ≤  (4 − ).
3 3

4. The function f (x) = (−1)x , x = 0, f (0) = 1 is undefined for some − < x <  for any
 > 0.
Remark: Notice that the statement limx→a f (x) = l contains ipso facto the implication
that f (x) exists in some neighborhood of x = a.
Finding the limit of f (x) as x → a can be looked upon as a process by which x is taken
closer and closer to a with the intention of discovering the ultimate value of f (x).
Definition of the limit according to Heine. Real number l is said to be the limit of
function f (x) at point a if to every sequence {xn }∞n=1 tenting to a, and for which f (xn ) is
defined for every xn , it happens that f (xn ) → l as n → ∞.
Theorem: The definitions of the limit according to Chauchy and Heine are equivalent.
Proof. Let l be the limit of function f (x) as x → a according to Chauchy. Let {xn }∞ n=1 tend
to a, with f (xn ) defined for any x, but with xn = a. According to Chauchy confinement of x
to a sufficiently small neighborhood of a assures that |f (x) − l| <  for 0 <  < 0 . From the
assumption that xn → a it results that 0 < |x − xn | < δ for all n > N (δ) and any willfully
small δ. Hence, |f (xn ) − l| <  for all n > N , f (xn ) → l and l is the limit of f (x) as x → a
also according to Heine.
Suppose now that l is the limit of function f (x) as x → a according to Heine. Deny
that l is the limit of f (x) at point a according to Chauchy. Then number η > 0 exists
so that |f (x) − l| ≥ η for all x sufficiently close to a. Thus, for any natural number n a
corresponding number xn can be found so that 0 < |xn − a| < 1/n and |f (xn ) − l| ≥ η. As
n → ∞ xn → a and therefore according to Heine sequence {f (xn )}∞ n=1 should converge to
l. But this is impossible if η > 0 and therefore our denial is untenable and l is the limit of
f (x) also according to Cauchy.
Examples.
1. Let f (x) = |x|/x, x = 0. Trying to find limx→0 |x|/x we start with the sequence
x = −1, −1/2, −1/3, −1/4 . . ., which tends to zero from below, and for which we always
compute f (x) = −1, tempting us to believe that the limit is −1. But the sequence x =
1, 1/2, 1/3, 1/4 . . ., which tends to zero from above, yields f (x) = 1, tempting us to believe
that the limit is 1. Since the limit may not depend on the approach, we conclude that there
is no limit to |x|/x as x → 0.
2. Let f (x) = sin x. Trying to find limx→∞ f (x) we select the sequence x = 0, π, 2π, 3π . . .
for which we compute f (x) = 0, tempting us to believe that the limit is zero. But the
sequence x = π/2, 5π/2, 9π/2 . . . for which we compute f (x) = 1 tempts us to believe that
the limit is one. Since the limit must be independent of the way x → x0 we conclude that
limx→∞ sin x does not exist.
3. That limx→0 (1/x2 ) does not exist results from the fact that 1/x2 becomes ever bigger as
|x| becomes ever smaller.
4
Chapter 6
4. We have that
x2 − 4 (x − 2)(x + 2)
lim = lim = lim (x + 2) = 4,
x→2 x − 2 x→2 x−2 x→2

even though (x2 + 4)/(x − 2) is not defined at x = 2.

2δx + |δx|
5. Find lim . (Ans. 0).
δx→0 1 + δx
δx + 2|δx|
6. Find lim . (Ans. none).
δx→0 δx
2 + 3|δx|
7. Find lim . (Ans. 2/3).
δx→0 3 − 5δx
|δx| − δx
8. Find lim . (Ans. none).
δx→0 δx + 2|δx|
|δx| δx
9. Find lim . (Ans. 0).
δx→0 (δx)2 + |δx|
|δx| δx + δx
10. Find lim . (Ans. 1).
δx→0 (δx)2 + |δx|
|δx| δx + δx
11. Find lim . (Ans. none).
δx→0 (δx)2 + δx
|δx| δx
12. Find lim . (Ans. none).
δx→0 |δx| + δx

Definition. We say that function f (x) → ∞ as x → a if by restricting x to a sufficiently

small neighborhood of a, function f (x) can be made willfully large.
For example
1
lim = ∞.
x→1 (x − 1)2

Indeed, if  
1 1
1− <x<1+
p p
then (x − 1)−2 > p for any p > 1.
Definition. We say that function f (x) → l as x → ∞ if by taking x sufficiently large
function f (x) can be brought willfully close to l.
For example
1−x
lim = −1.
x→∞ 1+x
Indeed, if x > (2 − )/, then
1−x
−1 < < −1 + 
1+x
for any 0 <  < 1.
Exercises.
5
 Chapter 6
1. Prove that there is no limit to f (x) = sin(1/x) as x → 0.
2. Prove that the limit of xn , n > 0, as x → a > 0, is an .
Theorem: The limit of a function is unique.
Proof: Say l1 and l2 are two different limits of f (x) as x → x0 . By restricting x to a
proper neighborhood of x0 , f (x) can be restricted to a willfully small neighborhood of l1
that excludes l2 , and vice versa. End of proof.
Implicit in the statement f (x) → l as x → a is the fact that f (x) is limited in its variation
around l due to a restriction on the variation of x around a. Computation of the limit does
not require that function f (x) be defined at x = a and, even if defined there, it need not be
that f (a) = l. Suppose, though, that f (x) can be evaluated at x = a.
Theorem: If f (a) exists (is finite), and limx→a f (x) = l, then f (x) is bounded in some
neighborhood of a.
Proof: Confining x to a sufficiently small neighborhood of a assures that |f (x) − l| < 0 for
some 0 > 0, or that l − 0 < f (x) < f + 0 . Hence

|f (x)| ≤ max( |l − 0 |, |l + 0 |, |f (a)| )

for all such x. End of proof.

Theorem: If in some neighborhood of a the inequality

f1 (x) ≤ f (x) ≤ f2 (x)

holds for any x = a, and if limx→a f1 (x) = limx→a f2 (x) = l, then also limx→a f (x) = l
Proof. From
f1 (x) − l ≤ f (x) − l ≤ f2 (x) − l
it results that as f1 (x) − l → 0 and f2 (x) − l → 0, then also f (x) − l → 0. End of proof.
Theorem: limx→a f (x) exists if and only if for any willfuly small  > 0 there is a δ > 0
such that for any x, y satisfying

0 < |x − a| < δ and 0 < |y − a| < δ

it happens that
|f (x) − f (y)| < .

Proof.
1. The condition is sufficient. Suppose the hypotheses of the theorem true, and let {xn =
a}∞n=1 be a sequence converging to a. Since xn → a there exists a natural number N so that
for any n > N it happens that 0 < |xn − a| < δ. Hence, if m, n > N , then according tho the
assumption of the theorem
|f (xn ) − f (xm )| < 
and {f (xn )}∞
n=1 is a Chauchy sequence converging to a (unique) limit.

6
Chapter 6
2. The condition is necessary. Suppose that limx→a f (x) = l. Then for any willfuly small
 > 0 there corresponds a δ such that for any x satisfying 0 < |x − a| < δ it happens that
|f (x) − l| < . If x and y are such that 0 < |x − a| < δ and 0 < |y − a| < δ, then

|f (x) − f (y)| = |(f (x) − l) + (−f (y) + l)| ≤ |f (x) − l| + |f (y) − l| < 2.

End of proof.
6.3 Limits of Sums and Products of Functions
If limx→a f (x) = l, then limx→a (l+(f (x)−l)) = l. Hence limx→a f (x) = limx→a (l+e(x))
where e(x) = f (x) − l → 0 as x → a.
Lemma: If limx→a e(x) = 0, then also limx→a (ce(x)) = 0 for any constant c.
Proof. Since e(x) → 0 as x → a, a neighborhoods of x = a exist for which |e(x)| can be
made willfully small. Consequently |ce(x)| can also be made willfully small. End of proof.
Lemma: Let f (x) be bounded in some neighborhood of x = a, namely |f (x)| < B, on
some interval around x = a. Then limx→a f (x)e(x) = 0 if limx→a e(x) = 0.
Proof. Since |f (x)e(x)| = |f (x)||e(x)| < B|e(x)|, then according to the previous lemma if
e(x) → 0 as x → 0, then so does B|e(x)| and consequently so does f (x)e(x). End of proof.
Lemma:
1 1
1≤ ≤ 2 if 0 ≤  ≤ .
1− 2

Proof. The hypothesis that  ≤ 1/2 implies that 1 −  > 0. Multiplying both sides of the
inequality by 1 −  > 0 we verify it. End of proof.
Theorem: If the limits exist and limx→a f (x) = l1 and limx→a g(x) = l2 , then
lim (f + g) = l1 + l2 .
1. x→a
lim (f g) = l1 l2 .
2. x→a
1 1
3. x→a
lim = , provided l2 = 0.
g l2
f l1
lim = , provided l2 = 0.
4. x→a
g l2
Proof. Write f (x) = l1 + e1 (x), g(x) = l2 + e2 (x) where e1 (x) → 0 and e2 (x) → 0 as x → a.
Then
1. f + g = (l1 + l2 ) + (e1 + e2 )
2. f g = (l1 l2 ) + (l1 e2 + l2 e1 + e1 e2 ).
1 1 1 e2 1 e2
3. = = − = − 2
g l2 + e2 l2 l2 (l2 + e2 ) l2 l2 (1 + e2 /l2 )
and since l2 = 0 and e2 → 0 as x → a, 1/(1 + e2 /l2 ) is, according to the lemma, bounded
near x = a.
f 1
4. = f .
g g
7
 Chapter 6
End of proof.
Theorem: Let f (x) be defined in some neighborhood of x = a, possibly excluding a itself. If
limx→a f (x) = l > 0, then a neighborhood of a, excluding x = a exists at which f (x) > 0.
proof. A neighborhood of a, x = a, exists at which l −  ≤ f (x) ≤ l + , with a willfully
small  > 0. End of proof.
Theorem:
1. lim |f (x)| = | lim f (x)|.
x→a x→a
2. If f (x) ≥ 0 in the environs of x = a, then lim f (x) ≥ 0.
x→a
lim f (x) ≥ x→a
3. If f (x) ≥ g(x) in the environs of x = a, then x→a lim g(x).
6.4 Limit of Trigonometric Functions
Theorem: limx→0 sin x = 0. and limx→0 cos x = 1
proof. Looking at the figure to the right
 we see that, for
x > 0, 0 < sin x < x. Also, cos x = 1 − sin2 x. End of
proof
Theorem: limx→0 sin x/x = 1.
proof. Looking at the figure above we see that the
area of sector OAB is greater then the area of triangle D
OAB but less than triangle OCD. Or sin x cos x < x < B
sin x/ cos x so that
cos x < x/ sin x < 1/ cos x. 2cosx 2sinx 2tanx=2sinx/co

But cos x → 1 as x → 0. End of proof. x 2x

1 O 1 A C
6.5 One Sided Limits
Definition: limx→a+ f (x) = r is the right limit of f (x) if a neighborhood a < x < a +
δ, δ() > 0 exists to the right of a so that for every x in this neighborhood r− < f (x) < r+,
for any willfully small . Similarly, limx→a− f (x) = l is the left limit of f (x) as x tends to
a from the left.
√
Example. If f (x) = x, then limx→0+ f (x) = f (0) = 0.
Theorem: limx→a f (x) = l if and only if limx→a− f (x) = limx→a+ f (x) = l.
Proof. Obviously, if limx→a− f (x) = limx→a+ f (x), then limx→a f (x) does not exist. As-
sume that the right and left limits of the function are the same. Then on the right for any
willfuly small  > 0 there corresponds a δR > 0 such that for any x on a < x < a + δR for
which f (x) is defined |f (x) − l| < . Similarly on the left, for the same  there exists δL > 0
such that for any x on a − δL < x < a for which f (x) is defined |f (x) − l| < . Taking
δ = min(δR , δL ) we have that if |x − a| < δ, then

a − δL ≤ a − δ < x < a + δ ≤ a + δR

8
Chapter 6
and |f (x) − l| <  so that f (x) → l as x → a. End of proof.
Theorem: If function f (x) ≥ 0 for all x > a, then limx→a+ f (x) ≥ 0.
Theorem: Let f (x) be an increasing function in the open
interval a < x < b, namely f (x1 ) > f (x0 ) for any x1 > x0 in
the interval. If f (x) is bounded from above, with u being its
least upper bound, then limx→b− f (x) = u.
Proof. If u is the least upper bound on f (x), then f (x) ≤ u Least Upper Bound
for any x in the interval. Let  > 0 be willfully small. For u
any such , u −  is not an upper bound on f (x). Hence for
any c < x < b, u −  < f (x) < u. End of proof. See the
figure to the right. f(x)
Let variable x of function f (x) is restricted to the open
u-ε
interval a < x < b. We may give sense to f (a+ ) and f (b− ) by
designating them to be the values of f (x) for x immediately
to the right of x = a and immediately to the left of x =
b, respectively, namely f (a+ ) = limx→a+ f (x) and f (b− ) =
limx→b− f (x). Say f (x) = x, 0 < x < 1 so that 0 < f (x) < 1. a c x b
+ −
Then f (a ) = 0 and f (b ) = 1. We may also employ the curious notational device of saying
that f (a + 0) = 0 and f (b − 0) = 1.
limx→0+ 1/x = ∞ means that 1/x can be made willfully large by selecting a sufficiently
small x > 0. Similarly, limx→0− 1/x = −∞ means that |1/x| can be made willfully large by
selecting a sufficiently small |x|, x < 0. Thus, f (x) = 1/x is defined for any x ∈ (0, 1), but
f (0+ ) = ∞.
Exercises.
1. Find √ √
a2 + x − a2 − x
lim .
x→0 x
2. Find √ √
a2 + x2 − a2 − x 2
lim .
x→0 x
3. Find √ √
a2 + x − a2 − x
lim .
x→0 x2
4. Show that
1
if f (x) = , then lim f (x) = 0 and lim f (x) = 1.
1 + 21/x x→0
x>0
x→0
x<0

√
5. We know that limx→−1 5 − 4x = 3. Find the limitions on x if
√
3−< 5 − 4x < 3 + .

9
 Chapter 6
6. We know that
2 − 3x
lim = 2.
x→0 1 + 5x

Find the limitions on x if

2 − 3x
2−< < 2 + .
1 + 5x

10