Chapter 13

Cardinality

Sometimes these cogitations still amaze The troubled midnight, and the noons repose. T. S. Eliot (La Figlia Che Piange)

FRAZZ: c Jef Mallett / Dist. by United Feature Syndicate, Inc. Reprinted with permission.

Before You Get Started. The goal of this chapter is to compare the sizes of innite sets. It is perfectly sensible to say that the sets {1, 2, 4} and {2, 3, 5} have the same size (still you might think about how to dene rigorously what it means for two nite sets to have equal size), but how do we compare the sizes of N and Z, of N and R>0 , of Q and R? We want to say more than just they are all innitehow do they compare? Are they all of different sizes? More generally, how should we measure the size of innite sets?

M. Beck and R. Geoghegan, The Art of Proof: Basic Training for Deeper Mathematics, Undergraduate Texts in Mathematics, DOI 10.1007/978-1-4419-7023-7_13, Matthias Beck and Ross Geoghegan 2010

121

122

13 Cardinality

13.1 Injections, Surjections, and Bijections Revisited

Here is what we mean when we say that two sets have the same size: The sets A and B have the same cardinality (or have the same cardinal number) if there exists a bijection A B. A special case is that of nite sets, for example, {1, 2, . . . , n} for some n N. Since we will use this set frequently in this chapter, we denote {1, 2, . . . , n} by [n]. A set S is nite if either S = or for some n N there exists a bijection from [n] to S. An innite set is one that is not nite. A set S is countably innite if there exists a bijection from N to S. A set S is countable if either S is nite or S is countably innite. It may seem obvious that there is no bijection [m] [n] when m = n, but it needs proof (Theorem 13.4) and is not trivial. The steps needed for the proof are given here as Propositions 13.113.3. Proposition 13.1. There exists no bijection [1] [n] when n > 1.
In other words, f and f are dened by the same rule, but they have different domains and codomains.

Proposition 13.2. If f : A B is a bijection and a A, dene the new function f : A {a} B { f (a)} Then f is well dened and bijective. Proposition 13.3. If 1 k n then the function gk : [n 1] [n] {k} is a bijection. Theorem 13.4. Let m, n N. If m = n, there exists no bijection [m] [n]. Proof. We prove the statement P(m): For all n = m there exists no bijection [m] [n] by induction on m N. The base case is captured by Proposition 13.1. For the induction step, let m 2 and assume we know that P(m 1) is true. Suppose (by way of contradiction) there exists a bijection f : [m] [n] for some n = m. Let k = f (m) and dene, as in Proposition 13.2, f : [m 1] [n] {k} by f(x) := f (x) . dened by gk ( j) := j j+1 if j < k, if j k by f(x) := f (x) .

13.1 Injections, Surjections, and Bijections Revisited

123

Proposition 13.2 says that this new function is also bijective. The composition of f with the inverse of the function gk dened in Proposition 13.3 gives a bijection [m 1] [n 1], by Proposition 9.7. But this contradicts our induction hypothesis.

Thus for nite sets the number of elements is well dened: a set S contains n elements if and only if there exists a bijection from [n] to S. Then every set having the same cardinal number as S also contains n elements. We say that contains 0 elements. Proposition 13.5 (Pigeonhole Principle). If m > n then a function [m] [n] cannot be injective. Proposition 13.5 implies that if m > n and we label n objects with numbers from 1 to m then there exist two objects that have the same label. The Pigeonhole Principle appears in many different areas in mathematics and beyond. It asserts that if there are n pigeonholes and m pigeons, there are at least two pigeons who must share a hole; or if there are n people in an elevator and m buttons are pressed, someone is playing a practical joke. Proposition 13.6. Every subset of a nite set is nite. Proposition 13.7. A nonempty subset of N is nite if and only if it is bounded above. Proposition 13.8. N is innite. Proposition 13.9. The nonempty set A is countable if and only if there exists a surjection N A. Proof. Assume A is nonempty and countable. If A is nite, then there exist n N and a bijection : [n] A. Let : N [n] be dened by m if 1 m n, (m) := 1 otherwise. This function is surjective, and so by Proposition 9.7, : N A is a surjection. If A is innite, then there exists a bijection from N to A, which is certainly surjective. Conversely, assume there exists a surjection : N A. If A is nite, it is countable by denition, and we are done. If A is innite, we dene a bijection : N A recursively as follows: (1) = (1), and for n 2, (n) = (mn ) ,

124 The set on the right-hand side is not empty because A is innite.

13 Cardinality

where mn := min k N : (k) / { (1), (2), . . . , (n 1)} . Proposition 13.10. A subset of a countable set is countable. Proposition 13.11. Every innite set contains an innite subset that is countable. Theorem 13.12. A set is innite if and only if it contains a proper subset that is also innite.

13.2 Some Countable Sets

The next propositions are counterintuitive at rst sight. Proposition 13.13. Z is countable. Theorem 13.14. Z Z is countable. Here is the idea of the proof:

Project 13.15. Find an explicit formula for a bijection between N and Z Z. Corollary 13.16. N N is countable. Corollary 13.17. Z (Z {0}) is countable. Corollary 13.18. Q is countable. We know that N Z Q; moreover, each is a proper subset of the next one, i.e., N = Z and Z = Q. This might make you think that N is smaller than Z and that Z is

13.3 Some Uncountable Sets

125

smaller than Q. But we have just proved that these sets have the same cardinality i.e., they have the same size. This can be confusing for beginners: if A and B are nite sets and A is a proper subset of B, then A and B have different cardinality by Theorem 13.4. We are seeing here that no such statement holds for innite sets. Proposition 13.19. The countable union of countable sets is countable, i.e., if An is a countable set for each n N then n=1 An is countable. A real number is algebraic if it is the root of a polynomial with integer coefcients. A real number that is not algebraic is called transcendental. Example 13.20. Every rational number is algebraic. The irrational numbers 3 and 4 2 are algebraic. The numbers and e, which you have studied in trigonometry and calculus, are transcendental (but this is not easy to prove). Proposition 13.21. The set of algebraic numbers is countable.
Can you see why r Q and 3 are algebraic?

Hint: use Proposition 6.20.

13.3 Some Uncountable Sets

Theorem 13.22. R is not countable. Proof. We will prove this by contradiction. Suppose that R is countable, and so by Proposition 13.9 there exists a surjective function f : N R. By Theorem 12.8, every real number has at most two decimal representations. So for each n N, f (n) can be (n) (n) (n) written in the form m(n) .d1 d2 d3 . . . ; if there is more than one such decimal for f (n), we use the one that has innitely many nonzero digits. Now let y be the real number represented by 0.a1 a2 a3 . . . , where an := 3 4 if dn = 3, if dn = 3.
(n) (n)

Then for all n N, y = f (n), because the nth decimal places of y and f (n) do not agree. Hence y R is not in the image of f , which contradicts the fact that f is surjective. It follows that there is no one-to-one correspondence between the innite sets R and Q, i.e., no function f : Q R that is bijective. In particular, the inclusion function Q R that takes each rational number to itself (regarded as a real number) is not surjective. This gives another proof that there exist irrational real numbers. The discovery that R and Q have different cardinality, i.e., different size, was considered revolutionary in the mathematics of the late nineteenth century. It was not

This diagonalization argument is due to Georg Cantor (18451928). Its appearance (Theorem 13.22) was the rst proof that innite sets may have different cardinal numbers.

126 Cantor originally conjectured the Continuum Hypothesis: there is no uncountable set whose cardinality is smaller than that of R. Kurt G odel (19061978) and Paul Cohen (19342007) proved that the usual axioms of set theory do not imply or refute the Continuum Hypothesis.

13 Cardinality

that people had thought the opposite to be true; they just had never seriously considered the idea of innite sets having different sizes. The foundations of the part of mathematics called analysis had to be completely rethought because of this. Corollary 13.23. The set R Q of irrational numbers is uncountable. Corollary 13.24. The set of transcendental numbers is uncountable. The proof of Theorem 13.22 reveals even more. It shows that the set of decimals 0.d1 d2 d3 : each d j = 3 or 4 R is uncountable. Consequently, the interval [0, 1] = {x R : 0 x 1} is uncountable. This construction can be modied to prove the following theorem.

A similar statement holds for open and half-open intervals.

Theorem 13.25. Every interval [x, y] is uncountable. This gives us a new proof of Proposition 11.17: Corollary 13.26. Let x, y R with x < y. Then there exists an irrational number z such that x < z < y. Proof. In any interval, there are only countably many rational numbers, so there must be an irrational number. Corollary 13.27. Between any two real numbers lies and algebraic number and also a transcendental number. Every interval is uncountable, and R has larger cardinality than N. A natural question is, where does the cardinality of an interval t into this picture? Here is the answer: Theorem 13.28. Every open interval (a, b) has the same cardinality as R. Corollary 13.29. All open intervals have the same cardinality.

This proof establishes the existence of irrational numbers in all intervals without explicitly describing any, in contrast with our hint for Proposition 11.17.

13.4 An Innite Hierarchy of Innities

Note that we use the card symbol only when comparing the cardinalities of two sets. We discuss cardinal numbers further in Chapter F.

We write card A card B if there exists an injection A B. By Proposition 9.12, card A card B is equivalent to saying that there exists a surjection B A. We write card A = card B if A and B have the same cardinality, i.e., if there exists a bijection A B. We write card A < card B when card A card B and card A = card B. If A is a set, let P(A) denote the set containing all subsets of A, called the power set of A.

13.5 Nondescribable Numbers

127

Example 13.30. If A = {a, b}, P(A) = {, {a}, {b}, {a, b}} . In this example, A has 2 members, and P(A) has 4 members. Our goal in this section is to prove that the power set of A is always bigger than A: Theorem 13.31. For every set A, card A < card P(A). Theorem 13.31 is profound. It implies an innite hierarchy of innities. For example, it says that P(N) is not countable, P(P(N)) has larger cardinality than P(N), P(P(P(N))) is yet larger than P(P(N)), etc. We start the proof of Theorem 13.31 with the nite case, for which we can be more precise: Proposition 13.32. For each n N, card P([n]) = card [2n ] . Proof of Theorem 13.31. The injection : A P(A), (x) = {x} shows that card A card P(A); we show that there is no surjection A P(A). Suppose, by way of contradiction, that f : A P(A) is surjective. Let B := {a A : a / f (a)} . This set B is an element of P(A), so by our assumption, there exists c A such that f (c) = B. If c B then c f (c), and so c / B. If c / B then c / f (c), and so c B. Either way we arrive at a contradiction. Cardinality questions are often difcult to answer. For example, the CantorSchr oder Bernstein Theorem asserts that if card A card B and card B card A then card A = card B. Another important theorem says that if A and B are sets, then either card A card B or card B card A. These are proved in Chapter F.
Before proving Proposition 13.32, think through the case n = 1.

Compare this argument with Cantor diagonalization in our proof of Theorem 13.22.

13.5 Nondescribable Numbers

This section is somewhat nonrigorous but is intended to make you aware of how elusive real numbers are. We have proved that the set R of real numbers is uncountable. There is a sense in which many members of R cannot be described at all. We have already referred to algorithms and have admitted that dening the word algorithm is complicated.

128

13 Cardinality

Here, we will simply say that an algorithm is a nite set of mathematical procedures that when applied to an input produces an output. As an example, suppose you input the question: nd 313 times 498 into your calculator. The answer 155874 will appear on the screen: this is the output. Programmed into your calculator is a multiplication algorithm that works on your input and gives the output. If you use a large and powerful computer rather than a calculator, you can input huge numbers and get huge outputs, though it may take some time for the computer to execute all the instructions given by the algorithm. The important point to note is that the algorithms length is not related to the size of the numbers you input, but rather the number of steps in applying the algorithm to your input will depend on the size of the input. The algorithm is a program containing loops, and the computer uses these loops as often as necessary for a particular input. The question is this: given a real number x, is there an algorithm that will print out as many decimal places of x as you desire? In other words, will the algorithm print out arbitrarily many decimal places? What might we mean by given a real number?
Here is the ratio of the circumference of a circle to its diameter. This formula can be derived from the Taylor series of the function arctan.

Example 13.33. You learned in calculus that 1 1 1 = 1 + + . 4 3 5 7

j+1 1 Thus the sequence of partial sums k j=1 (1) 2 j1 the sequence of partial sums of k=1

converges to 4 , and so

j=1

(1) j+1 2 j 1

converges to . An algorithm that computes the decimal expansion of these partial sums can print out arbitrarily many decimal places of . You can say how accurate you want the answer to be; this can easily be translated into what partial sum must be calculated, and the algorithms built in to the computer will do the rest. Example 13.34. We dened 2 = sup x R : x2 < 2 . Thus an algorithm can be described that will compute successive approximations to 2. What we see from these two examples is that the numbers and 2 are describable by a nite set of ordinary symbols, namely the symbols appearing in the relevant algorithms. This brings up the question whether some real numbers are not describable in this sense. By an ordinary symbol we mean any symbol that a reader of this book would recognize: lowercase and uppercase Latin and Greek letters, digits, and other common symbols such as space, period, left parenthesis, etc. We will call any such symbol a letter.

13.5 Nondescribable Numbers

129

By a word we mean a nite sequence of letters. Examples of words are good grade which is a 10-letter word, and sup x R : x2 < 2 which is a 13-letter word. This book is also a word. Proposition 13.35. The set of words is countable. We call x R describable if there exist m N and an m-letter algorithm that computes the decimal expansion of x in the manner described above. Since there are only countably many words, there are only countably many algorithms. It follows that many real numbers are not describable. In fact, the set of nondescribable real numbers is uncountable. In the earlier history of mathematics, it was thought that there was a chasm separating the continuous from the discrete, or, if you like, R from Z. Gradually it became clear that all real numbers can be understood in terms of integers via decimals or suprema. But the chasm reappears in a more subtle way. While there exists a decimal representation for each real number, we now see that for most real numbers a decimal description cannot actually be written down. Project 13.36. Which of the axioms for R are satised by the set of all describable numbers? Review Question. In what sense is the set Z of integers smaller than the set R of real numbers?
Weekly reminder: Reading mathematics is not like reading novels or history. You need to think slowly about every sentence. Usually, you will need to reread the same material later, often more than one rereading. This is a short book. Its core material occupies about 140 pages. Yet it takes a semester for most students to master this material. In summary: read line by line, not page by page. In technical language, the describable numbers form an ordered eld. Remember space is a letter.