4-8-2008

Counting and Cardinality

The cardinality of a set is roughly the number of elements in a set. This poses few diculties with
nite sets, but innite sets require some care.
I can tell that two sets have the same number of elements by trying to pair the elements up. For
example, the sets
{a, b, c, d} and {1, 2, 3, Calvin}
have the same number of elements because I can pair the elements of the rst set with the elements of the
second:
a b c d

1 2 3 Calvin
This kind of pairing is called a bijection or a one-to-one correspondence; its easy to understand
with nite sets, but I need to be more careful if Im going to use the same idea with innite sets. Ill begin
by reviewing the relevant denitions.
Denition. Let X and Y be sets and let f : X Y be a function.
1. f is injective if f(x) = f(y) implies x = y.
2. f is surjective if for all y Y , there is an x X such that f(x) = y.
3. f is bijective (or a one-to-one correspondence) if it is injective and surjective.
Denition. Let S and T be sets, and let f : S T be a function from S to T. A function g : T S is
called the inverse of f if
g (f(s)) = s for all s S and f (g(t)) = t for all t T.
I proved the following lemma earlier.
Lemma. Let S and T be sets, and let f : S T be a function. f is invertible if and only if f is bijective.
Example. Let
S = {a, b, c, d} and T = {1, 2, 3, Calvin}.
Dene f : S T by
f(a) = 1, f(b) = 2, f(c) = 3, f(d) = Calvin.
Ill show that f is a one-to-one correspondence by constructing an inverse for f. The inverse should
undo the eect of f:
a a
b b
c c
d d
1
2
3
Calvin
f f
-1
1
As you can see, I need to dene
f
1
(1) = a, f
1
(2) = b, f
1
(c) = 3, f
1
(Calvin) = d.
Ive constructed f
1
so that f
1
(f(s)) = s for all s S. To be complete, I should check that it works
the other way, too:
f

f
1
(1)

= f(a) = 1, f

f
1
(2)

= f(b) = 2, f

f
1
(3)

= f(c) = 3,
f

f
1
(Calvin)

= f(d) = Calvin.
So f
1
really is the inverse of f, and f is a bijection. (For that matter, f
1
is a bijection as well,
because the inverse of f
1
is f.)
Notice that this function is also a bijection from S to T:
h(a) = 3, h(b) = Calvin, h(c) = 2, h(d) = 1.
If there is one bijection from a set to another set, there are many (unless both sets have a single element).
I introduced bijections in order to be able to dene what it means for two sets to have the same number
of elements. The number of elements in a set is called the cardinality of the set.
Denition. Sets S and T have the same cardinality if there is a bijection f from S to T. Ill write |S| = |T|
to mean that S and T have the same cardinality.
A set S is nite if it is empty, or if there is a bijection f : {1, 2, 3, . . . , n} S for some integer n 1.
A set which is not nite is innite.
If S is a nonempty nite set and there is a bijection f : {1, 2, 3, . . . , n} S for some integer n 1, Ill
say that S has cardinality n or that S has n elements. In this case, Ill write |S| = n.
At this point, there is an apparently silly issue that needs to be resolved: Could a nite set be bijective
with both {1, 2, 3} and {1, 2, 3, 4} (say)? Of course, everyday experience says that this is impossible. However,
mathematicians always take the point of view that if something is really obvious, then it ought to be easy
to justify.
Actually, this particular point isnt that simple to justify try to prove it yourself! but its true,
and Ill omit the proof.
Example. Prove that the set of natural numbers N = {1, 2, 3, 4, . . .} has the same cardinality as the set
E = {2, 4, 6, 8, . . .} of positive even integers.
Dene f : N E by
f(n) = 2n.
This function has an inverse f
1
: E N given by
f
1
(m) =
m
2
.
Note that since m E, m is even, so m is divisible by 2 and
m
2
is actually a positive integer.
Heres the proof that f and f
1
are inverses:
f

f
1
(m)

= f

m
2

= 2
m
2
= m for m E,
f
1
(f(n)) = f
1
(2n) =
2n
2
= n for n N.
2
This situation looks a little strange. E is contained in N, but Ive just shown that the two sets have the
same number of elements. The only reason this looks funny is that it contradicts your real world experience
which only deals with nite objects. In fact, its characteristic of innite sets that they have the same
number of elements as some of their proper subsets.
Informally, if the elements of an innite set can be listed
a
1
, a
2
, a
3
, . . . ,
then the set has the same cardinality as the natural numbers.
In fact, to dene listable precisely, youd end up saying the set has the same cardinality as the natural
numbers. But this is a good picture to keep in mind. Ill show that the real numbers, for instance, cant be
arranged in a list in this way.
The next part of this discussion points out that the notion of cardinality behaves the way the number
of things in a set ought to behave.
Lemma. Let S, T, and U be sets.
(a) The identity function id : S S given by id(s) = s is a bijection.
(b) The inverse of a bijection is a bijection.
(c) If f : S T and g : T U are bijections, then the composite g f : S U is a bijection.
Proof. (a) The identity function has an inverse, namely itself. Therefore, the identity function is a bijection.
(b) If f : S T is a bijection, then by denition it has an inverse f
1
: T S. To be inverses means that
f
1
(f(s)) = s for all s S and f

f
1
(t)

= t for all t T.
But these equation also say that f is the inverse of f
1
, so it follows that f
1
is a bijection.
(c) Suppose that f : S T and g : T U are bijections. Let f
1
: T S and g
1
: U T be their
respective inverses. Ill prove that f
1
g
1
is the inverse of g f.

f
1
g
1

(g f)

(s) = f
1

g
1
(g (f(s)))

Denition of composite
= f
1
(f(s)) g
1
(g (junk)) = junk
= s f
1
(f (junk)) = junk

(g f)

f
1
g
1

(u) = g

f
1

g
1
(u)

Denition of composite
= g

g
1
(u)

f
1
(junk)

= junk
= u g
1
(g (junk)) = junk
This proves that f
1
g
1
is the inverse of g f, so g f is a bijection.
Corollary. Let S, T, and U be sets.
(a) (Reexivity) |S| = |S|.
(b) (Symmetry) If |S| = |T|, then |T| = |S|.
(c) (Transitivity) If |S| = |T| and |T| = |U|, then |S| = |U|.
Proof. (a) By the lemma, the identity function id : S S is a bijection, so |S| = |S|.
(b) If |S| = |T|, then there is a bijection f : S T. By the lemma, f
1
: T S is a bijection. Therefore,
|T| = |S|.
3
(c) If |S| = |T| and |T| = |U|, then there are bijections f : S T and g : T U. By the lemma,
g f : S U is a bijection, so |S| = |U|.
As you know, these three properties mean that having the same cardinality is an equivalence relation.
This actually justies my use of the = when I write |S| = |T|.
Example. Prove that the interval (0, 1) has the same cardinality as R.
(Remember that (0, 1) stands for the open interval {x R | 0 < x < 1}.)
First, notice that the open interval

2
,

2

has the same cardinality as the real line. To prove this, I

have to construct a bijection f :

2
,

2

R. Its easy: just dene

f(x) = tanx.
To show that f is bijective, I have to show that it has an inverse; the inverse is f
1
(x) = arctanx.
Now I know that

2
,

2

and R have the same cardinality. Next, Ill show that (0, 1) and

2
,

2

have the same cardinality.

The idea is to multiply by to stretch (0, 1) to (0, ). Then I subtract

2
to shift (0, ) to

2
,

2

.
1
0
p
p/2 -p/2
p multiply by
p/2 subtract
All together, I dene g : (0, 1)

2
,

2

by
g(x) = x

2
.
First, if 0 < x < 1, then 0 < x < , so

2
< x

2
<

2
. This shows that g takes inputs in (0, 1)
and produces outputs in

2
,

2

.
To show that g is bijective, I have to produce an inverse. The standard swap the xs and ys procedure
works; you get
g
1
(x) =
x +

2

.
Heres the proof that g and g
1
are inverses:
g

g
1
(x)

= g

x +

2

=
x +

2

2
= x +

2


2
= x,
g
1
(g(x)) = g
1

x

2

=
x

2
+

2

=
x

= x.
4
Therefore, g is a bijection, so (0, 1) and

2
,

2

have the same cardinality. By transitivity, (0, 1) and

R have the same cardinality.
Example. Prove that (0, 1) has the same cardinality as R
+
= (0, ).
Dene f : (0, 1) (1, ) by
f(x) =
1
x
.
Note that if
0 < x < 1, then
1
x
> 1.
Therefore, f does map (0, 1) to (1, ).
0 1
f(x) = 1/x swaps these intervals
I claim that f
1
(x) =
1
x
. If x > 1, then 0 <
1
x
< 1, so f
1
maps (1, ) to (0, 1). Moreover,
f

f
1
(x)

= f

1
x

=
1

1
x
= x,
f
1
(f(x)) = f
1

1
x

=
1

1
x
= x.
Thus, f is a bijection.
Dene g : (1, ) (0, ) by
g(x) = x 1.
If x > 1, then x 1 > 0. Therefore, g does map (1, ) to (0, ).
I claim that g
1
(x) = x + 1. If x > 0, then x + 1 > 1, so g
1
maps (0, ) to (1, ). Moreover,
g

g
1
(x)

= g(x + 1) = (x + 1) 1 = x,
g
1
(g(x)) = g
1
(x 1) = (x 1) + 1 = x.
Therefore, g is a bijection.
With the bijections f and g, I have |(0, 1)| = |(1, )| = |(0, )|, so (0, 1) and (0, ) have the same
cardinality.
In many situations, its dicult to show that two sets have the same cardinality by actually constructing
a bijection between them. The theorem that follows gives an indirect way to show that two sets have the
same cardinality.
Theorem. (Schr oder-Bernstein) Let S and T be sets. Suppose there are injective functions f : S T and
g : T S. Then S and T have the same cardinality.
The proof of the Schr oder-Bernstein theorem is a little tricky, so I wont do it here.
5
The Schr oder-Bernstein theorem says that if S has the same cardinality as a subset of T, and T has the
same cardinality as a subset of S, then S and T must have the same cardinality.
S
S
T
T
It is a powerful tool for showing that sets have the same cardinality. Here are some examples.
Example. Show that the open interval (0, 1) and the closed interval [0, 1] have the same cardinality.
The open interval 0 < x < 1 is a subset of the closed interval 0 x 1. In this situation, there is an
obvious injective function f : (0, 1) [0, 1], namely the function f(x) = x for all x (0, 1). (f is called
an inclusion map.) If f(x
1
) = f(x
2
), then x
1
= x
2
, so f is injective.
Next, Ill construct an injective function g : [0, 1] (0, 1). The idea is to nd a copy of [0, 1] in (0, 1),
then do some scaling and translation to map [0, 1] onto the copy. Ill use the interval [0.25, 0.75] as my target
in (0, 1). The target has length 0.5, so Ill multiply by 0.5 to shrink [0, 1] to [0, 0.5]. Next, Ill add 0.25 to
shift [0, 0.5] to [0.25, 0.75]. All together, I get
g(x) = 0.5x + 0.25 for 0 x 1.
First, if 0 x 1, then 0 0.5x 0.5, so 0.25 0.5x + 0.25 0.75. This proves that g is a function
from [0, 1] to [0.25, 0.75].
Next, I need to show that g is injective. Suppose that g(x
1
) = g(x
2
), I must prove that x
1
= x
2
. Now
g(x
1
) = g(x
2
) means that
0.5x
1
+ 0.25 = 0.5x
2
+ 0.25, so 0.5x
1
= 0.5x
2
, and hence x
1
= x
2
.
Therefore, g is injective. (In fact, g is bijective, and you could prove injectivity by constructing g
1

though it would be overdoing it a bit.)

Now I have injective functions (0, 1) [0, 1] and [0, 1] (0, 1). By Schr oder-Bernstein, |(0, 1)| = |[0, 1]|.
Example. Prove that [1, 1] has the same cardinality as (1, 3) (4, 6).
-1 1
1 3 4 6
The two sets dont look alike the rst set is a single interval which is closed on both ends, while
the second set consists of two open intervals. When two sets dont look alike but you think they have the
6
same cardinality, consider using the Schr oder-Bernstein theorem. Ill dene injective functions going from
each set into the other.
Ill describe in words how Im getting the denitions of the functions. (Note that there are many
functions you could use to do this!)
The rst set is an interval of length 2, which (because of its endpoints) wont t in either of the intervals
that make up the second set. Since the second sets intervals dont have endpoints, if I just slide [1, 1] over,
its endpoints will stick out of the ends of either (1, 3) or (4, 6). So the idea is to shrink [1, 1] rst, then
slide it inside either (1, 3) or (4, 6).
If I multiply [1, 1] by 0.5, I get [0.5, 0.5], an interval of length 1. This will surely t inside (1, 3)
(say), and I can slide [0.5, 0.5] into (1, 3) by adding 2. This takes [0.5, 0.5] to [1.5, 2.5]. I just have to do
the two steps one after the other. So dene f : [1, 1] (1, 3) (4, 6) by
f(x) = 0.5x + 2.
First, I have to show that this makes sense that is, that f really takes [1, 1] into (1, 3) (4, 6).
Suppose x [1, 1]. Then
1 x 1
0.5 0.5x 0.5
0.5 + 2 0.5x + 2 0.5 + 2
1.5 f(x) 2.5
Since 1.5 f(x) 2.5, obviously 1 < f(x) < 3, so f does map [1, 1] into (1, 3) (4, 6).
Next, I have to show that f is injective. Suppose f(a) = f(b). Then
0.5a + 2 = 0.5b + 2
0.5a = 0.5b
a = b
Hence, f is injective.
Next, I have to dene an injective function g : (1, 3) (4, 6) [1, 1]. Now (1, 3) (4, 6) occupies a
total length of 61 = 5, whereas the target interval [1, 1] has length 2. If I multiply by
1
5
= 0.2, Ill shrink
(1, 3) (4, 6) to (0.2, 0.6) (0.8, 1.2), which has a total length of 1. Next, I can slide (0.2, 0.6) (0.8, 1.2)
inside [1, 1] by subtracting 0.7, which should give (0.5, 0.1) (0.1, 0.5).
Thus, dene g : (1, 3) (4, 6) [1, 1] by
g(x) = 0.2x 0.7.
I need to check that g maps (1, 3) (4, 6) into [1, 1]. Let x (1, 3) (4, 6). Then certainly x is between
1 and 6, i.e. 1 < x < 6. (Of course, 1 < x < 6 does not imply that x (1, 3) (4, 6). Do you see why?) So
1 < x < 6
0.2 < 0.2x < 1.2
0.2 0.7 < 0.2x 0.7 < 1.2 0.7
0.5 < g(x) < 0.5
Since 0.5 < g(x) < 0.5, obviously 1 g(x) 1, so g does map (1, 3) (4, 6) into [1, 1].
Next, I have to show that g is injective. Suppose g(a) = g(b). Then
0.2a 0.7 = 0.2b 0.7
0.2a = 0.2b
a = b
7
Therefore, g is injective.
Hence, [1, 1] and (1, 3) (4, 6) have the same cardinality, by the Schr oder-Bernstein theorem.
Ive already noted that its easy to nd nite sets of dierent cardinalities: for example, a set with
three elements does not have the same cardinality as a set with 42 elements. Ive also given examples of
innite sets which have the same cardinality. Its an important fact that not all innite sets have the same
cardinality there are dierent kinds of innity! Heres some terminology which Ill used to describe the
situation.
Denition. A set is countably innite if it has the same cardinality as the natural numbers N =
{1, 2, 3, . . .}. An innite set which is not countably innite is uncountably innite or uncountable.
A set is countable if it is either nite or countably innite.
I know that some innite sets the even integers, for instance are countably innite. I know of
other innite sets, such as the real numbers. Is the set of real numbers countably innite? The answer is
no; the proof is due to Georg Cantor (18451918), and is called the diagonalization argument.
Theorem. The open interval (0, 1) is uncountably innite.
Proof. Im going to be a little informal in this proof so that the main idea isnt lost in a lot of notation.
Suppose on the contrary that (0, 1) is countably innite. Represent numbers in the interval as decimals
0.a
1
a
2
a
3
. . .. (If a number ends in an innite sequence of 9s, rewrite it as a nite decimal so, for instance,
0.134999 . . . becomes 0.135.) Since (0, 1) is countably innite by assumption, I can arrange the numbers in
(0, 1) in a list:
.
.
.
.
.
.
.
.
.
.
.
3 6 8 9 7 3 4 8
5 0 4 1 8 5 6 3
0 2 4 7 3 5 9 6
2 1 7 6 1 4 0 7
4 4 2 0 5 9 3 1
7 9 6 9 3 2 1 5
7 1 8 1 8 0 4 2
1 3 5 4 6 7 6 3
. . .
I emphasize that, by assumption, this list contains all of the numbers in the interval (0, 1).
Now go down the diagonal and make a number using the digits. In this case, I get the number
0.30465243 . . ..
Take each of the digits in this number and change it to any other digit except 9. For example, you could
add 1 to each digit from 0 to 7 and change 8 or 9 to 0. This would produce the number 0.41576354 . . ..
(The reason you do not want to change digits to 9 is so that you dont wind up with a number that
ends in an innite sequence of 9s.)
The number 0.41576354 . . . diers from each of the numbers in my list. Specically, the n
th
digit of
0.41576354 . . . is dierent from the n
th
digit in the n
th
number on the list. This means that 0.41576354 . . .
is not in my list which is a contradiction, because I assumed that my list contained all of the numbers in
the interval (0, 1).
Therefore, the interval (0, 1) must be uncountably innite.
Since the interval (0, 1) has the same cardinality as R, R is uncountably innite as well. Notice that Z
(which is countably innite) is a subset of R. Are there any sets which are between Z and R in cardinality?
The Continuum Hypothesis states that there are no sets which are between Z and R in cardinality;
it was rst stated by Cantor, who was unable to construct a proof. Kurt Godel [2] proved around 1940 that
the Continuum Hypothesis was consistent relative to the standard axioms of set theory. Paul Cohen [1]
8
proved in 1963 that the Continuum Hypothesis is undecidable: It is independent of the standard axioms for
set theory.
In other words, the question of the existence of a subset of R which has cardinality dierent from either
Z or R cant be settled without adding assumptions to standard mathematics and you can assume either
that such a set exists, or that it doesnt, without causing a contradiction.
Denition. Let S be a set. The power set P(S) of S is the set of all subsets of S.
Example. Let S = {a, b, c}. The power set of S is
P(S) = {, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}}.
Notice that the power set includes the empty set and the set S itself.
If youre constructing a subset of a set, there are two alternatives for each element: Either it is in the
subset, or it is not. So if the set has n elements, the two alternatives for each element give 2
n
possibilities
in all. Therefore, if S is nite and |S| = n, then | P(S)| = 2
n
.
In this example, |S| = 3, and | P(S)| = 2
3
= 8.
Theorem. If S is a set, then S and P(S) do not have the same cardinality.
Proof. Suppose rst that S = . Now , so P() = {}. Hence, || = 0 while |P()| = 1, and the result
is true in this case.
Now suppose that S = . Ill prove the result by contradiction. Suppose that |S| = | P(S)|. This means
that there is a bijection f : S P(S).
Since f is a bijection, every element of the power set that is, every subset of S is paired up with
an element of S. For example, there must be an element s S for which f(s) = .
Of course, s / . So s is an element which is paired up with a subset that doesnt contain it. And in
general, f takes an element of S to a subset of S, and that subset either contains the element or it doesnt.
Heres a particular example to help you get your bearings. In the picture below, the set is S = {a, b, c, d}
and the function f is depicted by the arrows.
f
{a}
{b}
{c}
{a,b}
{a,c}
{d}
{a,d}
{b,c}
{b,d}
{c,d}
{a,b,c}
{a,b,d} {a,c,d}
{b,c,d}
{a,b,c,d}
P(S)
a
d
c
b
S
f
In this example, f takes b and c to subsets that contain them; f takes a and d to subsets which dont
contain them.
Continuing with the proof, let
T = {s S | s / f(s)}.
9
That is, T is the subset of elements of S which f takes to subsets which dont contain them. I know
there is at least one such element, namely the element which f takes to the empty set.
Now f is bijective, and T is a subset of S, so there is an element s
0
S such that f(s
0
) = T. Question:
Is s
0
T?
If s
0
T, then by denition of T, s
0
/ f(s
0
) = T. This is a contradiction.
If s
0
/ T, then s
0
/ T = f(s
0
), so s
0
satises the dening condition for T which means s
0
T. This
is a contradiction.
Since s
0
T and s
0
/ T both lead to contradictions, Ive actually contradicted my rst assumption
that |S| = | P(S)|. Therefore, |S| = | P(S)|, as I wanted to prove.
Example. The power set of the natural numbers N has the same cardinality as R.
I showed earlier that N is countably innite, whereas R is uncountably innite, so this conrms the
theorem in this particular case.
Example. If S and T are nite, then |S T| = |S||T|.
For example, if S has 42 elements and T has 5 elements, then S T has 42 5 = 210 elements.
Interesting things happen when S and T are innite. For example, N is countably innite; how big is
N N? It turns out the N N is countably innite as well.
N N is the set of pairs (m, n), where m and n are natural numbers:
(1,1)
(1,2)
(1,3)
(1,4)
(2,1)
(2,2)
(2,3)
(2,4)
(3,1)
(3,2)
(3,3)
(3,4)
(4,1)
(4,2)
(4,3)
(4,4)
Im going to list the pairs starting with (1, 1) in the order shown by the grey line. This means Im
constructing a function f : N N N. Here it is:
f(m, n) =
(m+n 2)(m+n 1)
2
+m.
Here is why this works. (m, n) is the m
th
element on the diagonal line whose elements add up to m+n.
Previous to that, Ive gone through
0 + 1 + 2 + + (m+n 2) =
(m+n 2)(m+n 1)
2
elements. That gives
(m+n 2)(m+n 1)
2
+m.
Its a little tricky to show f is injective, so Ill omit the proof here. There is an obvious way to make an
injective function from N to N N:
g(n) = (1, n).
If g(n
1
) = g(n
2
), then (n
1
, 1) = (n
2
, 1), so n
1
= n
2
, and hence g is injective. By the Schr oder-Bernstein
theorem, N and N N have the same cardinality.
10
[1] Paul J. Cohen, Set Theory and the Continuum Hypothesis, Reading, Massachusetts: The Benjamin-
Cummings Publishing Company, Inc., 1966 [ISBN 0-8053-2327].
[2] Kurt Godel, Consistency-proof for the generalized continuum hypothesis, Proc. Nat. Acad. Sci. U.S.A.,
25(1939), 220-204.
c 2008 by Bruce Ikenaga 11