Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

A Proof That R is Uncountable Using the Intermediate

Value Theorem

It’s widely known that the set of real numbers is uncountable. We will give a proof
of this fact by showing that if the set of reals is countable, then the intermediate value
theorem fails to hold. Since the intermediate value theorem is a consequence of the axioms
of the real numbers, R must be uncountable. We’ll explain the prerequisite results, and
then give the proof at the end of this paper.

1. Enumerating Certain Totally Ordered Sets

Some notation and terms need to be clarified.


Remark. If (X, ≤) is a totally ordered set, we say x < y for x, y ∈ X whenever x ≤ y and
x 6= y. This defines a strict total order relation, which will be universally denoted as
“<” in this paper. We say (X, ≤) is orderwise dense if for any x, y ∈ X such that x < y,
there exists another element z ∈ X such that x < z < y.
Remark. When we have multiple indices, we will use concatenation, so instead of writing
xi,j we’ll write xij . This will be important to keep in mind when the indices will be more
elaborate expressions. For example, we will write x(n−1),2n−1 as x(n−1)2n−1 .

For the rest of this section, let (X, ≤) be any totally ordered set such that

(i) it is orderwise dense,


(ii) it is countably infinite, and
(iii) it has a maximum and a minimum.

We will show that these kinds of sets can have their elements labeled in a way that reveals
all these sets to be order-isomorphic to each other.
We’re assuming X has a minimum and a maximum. For convenience we will denote
the minimum by “0” and the maximum by “1.” Since X is a countably infinite set, we
can consider an arbitrary enumeration of the elements between the max and min: X =
{0, 1} ∪ {x1 , x2 , . . .}. Given this, we relabel the elements of X as follows. In the first step,
define
x10 := 0, x11 := x1 , x12 := 1, and X1 := {x10 , x11 , x12 }.
Note that X1 contains x1 , and x10 < x11 < x12 .
After the first step, the steps follow an inductive procedure. For n ≥ 2, the nth step
goes as follows. We assume we have the subset Xn−1 = {x(n−1)0 , x(n−1)1 , . . . , x(n−1)2n−1 }
of X such that

(1) x(n−1)0 = 0 and x(n−1)2n−1 = 1,


(2) x(n−1)0 < x(n−1)1 < · · · < x(n−1)2n−1 , and
(3) Xn−1 contains the elements x1 , . . . , xn−1 .

1
If 0 ≤ j ≤ 2n−1 , we define xn(2j) := x(n−1)j . This defines xnk for all even k inclusively
between 0 and 2n . A restatement of (2) in our inductive hypothesis says

xn0 < xn2 < xn4 < · · · < xn(2n −2) < xn2n . (∗)

Next, we define xnk for odd k between 0 and 2n . Fix such a k, for which k = 2j + 1. The
fact that 1 ≤ k ≤ 2n − 1 (k is odd so it can’t be zero or 2n ) implies 0 ≤ j ≤ 2n−1 − 1. Given
such a j, we can consider the interval (x(n−1)j , x(n−1)(j+1) ). Because of (2) and the fact that
X is orderwise dense, such an interval is nonempty. As such, let xi ∈ (x(n−1)j , x(n−1)(j+1) )
be an element of X for the least possible index i. Then define xnk := xi . This defines the
elements xnk for the rest of k’s inclusively between 0 and 2n , and we define

Xn := {xn0 , xn1 , . . . , xn2n }.

Now it is left to show that (1)-(3) hold with n in place of n − 1.

(1) Obviously, xn0 = x(n−1)0 = 0 and xn2n = x(n−1)2n−1 = 1 by definition.


(2) The definition of xnk for odd k and (∗) together imply

xn0 < xn1 < xn2 < · · · < xn(2n −1) < xn2n .

(3) The definition of xnk for even k implies Xn−1 ⊆ Xn . This immediately implies
x1 , . . . , xn−1 ∈ Xn , and it also implies xn ∈ Xn assuming xn is included in the old set
Xn−1 . If xn is not contained in the old set Xn−1 , then xn is not an endpoint of the any
of the intervals (x(n−1)0 , x(n−1)1 ), (x(n−1)1 , x(n−1)2 ), . . . , and (x(n−1)(2n−1 −1) , x(n−1)2n−1 ),
so it must be included in one of these intervals, say (x(n−1)j , x(n−1)(j+1) ). All elements
x1 , . . . , xn−1 with smaller index are already included in the endpoints of these inter-
vals (by (3)), so xn has the least possible index to be contained in (x(n−1)j , x(n−1)(j+1) ),
so xn ∈ Xn . This shows Xn contains x1 , . . . , xn .

This completes the inductive step.

The construction of the labeling system has a few important properties that we’ll need
to cite.

Proposition 1. For any n ≥ 1, we have

xn0 = 0 < xn1 < xn2 < · · · < xn(2n −1) < 1 = xn2n .

Proposition 2. For any n ≥ 1, we have xnk = x(n+1)(2k) for all 0 ≤ k ≤ 2n . Consequently,


we have X1 ⊆ X2 ⊆ X3 ⊆ · · · .
S
Proposition 3. We have X = n≥1 Xn , so our construction relabels all elements of X.

The first two propositions are direct consequences of our construction. For the last propo-
sition, remember that S {0, 1} ∪ {x1 , . . . , xn } ⊆ Xn ⊆ X, so by taking the union over all
n ≥ 1, we obtain X ⊆ n≥1 Xn ⊆ X.

2
2. Constructing the Order-Isomorphism

Suppose (X, ≤) and (Y, ≤) are two totally ordered sets such that they are orderwise dense,
they are
S countably infinite, and each comes withSa maximum and minimum. Then let
X = Xn with Xn = {xn0 , . . . , xn2n } and Y = Yn with Yn = {yn0 , . . . , yn2n } be the
labeling systems as described above.
We define f : X → Y as follows. Given x ∈ X, Proposition 3 guarantees that it is of
the form x = xnk for some n ≥ 1 and 0 ≤ k ≤ 2n , and we set

f (xnk ) = ynk .

It needs to be proven that f is well-defined. Suppose xnk = xn0 k0 for some n, n0 ≥ 1,


0
0 ≤ k ≤ 2n , and 0 ≤ k 0 ≤ 2n . Without loss of generality, assume n0 ≥ n (if not, switch the
relevant labels). By Proposition 2, we have xnk = xn0 (k2n0 −n ) . By Proposition 1, we are
0
forced to have k 0 = k2n −n . Then

ynk = yn0 (k2n0 −n ) = yn0 k0 ,

and so the output of f is independent of the choice of the label used. This construction
leaves f with a few notable properties.
Proposition 4. f is an order-isomorphism.
Proof. Let x, x̃ be two distinct elements of X. We may assume x < x̃. By Proposition 2 and
Proposition 3, we know x, x̃ ∈ Xn for sufficiently large n. For any such n, we have x = xnk
and x̃ = xnk0 for some integers k, k 0 inclusively between 0 and 2n . By Proposition 1, we
must have k < k 0 so then f (xnk ) = ynk < ynk0 = f (xnk0 ). This proves f is injective and
preserves order.
It is only left to show that f is surjective. But that easily follows if we apply Proposi-
tion 3 to Y . Any element y ∈ Y is of the form y = ynk so f (xnk ) = ynk . 

Proposition 5. f is continuous.
Remark. If (X, ≤) is a totally ordered set, the assumed topology is the topology generated
by open intervals (x, y) ⊆ X.

Proof. Let (y, y 0 ) be an open interval of Y . Obviously, f −1 ((y, y 0 )) = {x ∈ X | y < f (x) <
y 0 }. Since f is an order-isomorphism, the criteria that y < f (x) < y 0 is equivalent to
f −1 (y) < x < f −1 (y 0 ), so f −1 ((y, y 0 )) = (f −1 (y), f −1 (y 0 )). Thus the inverse image of an
open interval is an open interval. 

3. Proving the Intermediate Value Theorem

To make our main result as clear as possible, we will show that the intermediate value
theorem for the reals is a direct consequence of the least upper bound property of the real
numbers.
Theorem (The Intermediate Value Theorem). Suppose f : [x, y] → R is a continuous
function, and suppose c is a real number inclusively between f (x) and f (y). Then c is in
the image of f .

3
Proof. Suppose f (x) ≤ f (y) and c ∈ [f (x), f (y)]. Then let S = {z ∈ [x, y] | f (z) ≤ c} and
z0 = sup S. Obviously, x ∈ S and S ⊆ [x, y] so the set is nonempty and bounded; this
guarantees z0 exists as a real number.
Let ε > 0. Continuity says that there exists δ > 0 such that

z ∈ (z0 − δ, z0 + δ) =⇒ f (z) ∈ (f (z0 ) − ε, f (z0 ) + ε). (∗)

By definition of z0 , there exists z ∈ (z0 − δ, z0 ] ∩ S where f (z) ≤ c. This and (∗) imply
f (z0 ) − ε < f (z) ≤ c. Hence, f (z0 ) − ε ≤ c. Since ε > 0 is arbitrary, it follows that

f (z0 ) ≤ c.

If z0 = y, then the fact that c ≤ f (y) and the above inequality together imply c = f (z0 ),
proving the theorem. Hence, we’ll assume z0 < y. For any ε > 0, let δ > 0 such that (∗)
holds again. By definition of z0 , for any z ∈ (z0 , z0 + δ) ∩ [x, y] we have f (z) > c (such a z
exists because we’re assuming z0 < y). After choosing such a z, the fact that f (z) > c and
(∗) imply f (z0 ) + ε > f (z) > c. Hence f (z0 ) + ε > c. Since ε > 0 is arbitrary, it follows
that
f (z0 ) ≥ c.
Combining our obtained inequalities shows f (z0 ) = c, so c is in the image of f .
The proof for the case where f (x) ≥ f (y) is similar, so we omit it. 

4. Proving the Main Theorem

Now with the prerequisite theorems clarified, we are in a position to prove our main result.

Theorem. R is uncountable.

Proof. It suffices to prove that [0, 1] is uncountable. Suppose to the contrary. Then [0, 1]
is a totally ordered set such that (i) it is orderwise dense, (ii) it is countably infinite, and
(iii) it has a maximum and a minimum. Since [0, 1] ∩ Q has those same properties, there
is a continuous increasing bijection f : [0, 1] → [0, 1] ∩ Q where the domain and codomain
have the order topologies. Since f is increasing and onto [0, 1] ∩ Q, we must have f (0) = 0
and f (1) = 1.
Now it is known that [0, 1]∩Q with the order topology is a subspace of R, so the inclusion
map ι : [0, 1] ∩ Q ,→ R is continuous. The map g = ι ◦ f : [0, 1] → R is a composition
of continuous maps, so g is a continuous real function. Now g(0) = 0, g(1) = 1, and g is
continuous,
√ but the image of g is [0, 1] ∩ Q, so any irrational between g(0) and g(1) such
as 1/ 2 is not in the image of g. This contradicts the intermediate value theorem, and so
our initial supposition that [0, 1] is countable is false. 

You might also like