You are on page 1of 16

Peter Suber, "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber,%20Infinite%20Sets.

htm

Published in the St. John's Review, XLIV, 2 (1998) 35-59. Copyright © 1998, Peter
Suber.

This crash course is designed to stand alone. But it also functions as the appendix to
my essay, Infinite Reflections.

A Crash Course in the Mathematics
Of Infinite Sets
Peter Suber, Philosophy Department, Earlham College

Don't be surprised if this is easier than you thought. Set theory requires no algebra or
calculus. It is much more primitive than those branches of mathematics, and rests on
very simple notions. Moreover, the proofs will be unusually short and uncomplicated.

What will be difficult? Most of the results we will prove depend critically on those
that came before; but I cite the needed prior theorems by number to make this kind
of back-tracking easier. The notation may be new, and for many people unfamiliar
notation raises the hair on the back of the neck. But most of the notation may be
ignored. I include it mainly so that if you read further on this subject, you will be
equipped. I honestly don't think the compressed exposition needed for a crash course
increases the difficulty —in part because I've been more long-winded than most
mathematicians, and in part because some compression and conciseness helps keep
all the relevant ideas in the head at the same time, which aids comprehension. Some
of the proofs, short and simple as they are, will make you dizzy. But that's part of the
amazing phenomenon to be savored, not a difficulty to lament.

To begin:

Almost a definition. Intuitively, a set is a collection of elements.

The intuitive notion of a set leads to paradoxes, and there is considerable
mathematical and philosophical disagreement about how best to refine the intuitive
notion. Fortunately, none of the disagreements or refinements matters for our
purposes here. I only bring up this complexity so that you'll accept the intuitive notion
in place of a refined definition for the purposes of this crash course.
Notation. When we want to list the members of a set, we use curly brackets. So if
set S contains elements A, B, and C, then we say S = {A, B, C}.
The null set is the empty set or the set with no members. Notation: Ø. Hence, Ø =
{}.

Abbreviation. For if and only if I will sometimes write simply iff.

Definition. Set A is a subset of set B iff all the members of A are also members of B.

Notation. A B.
It follows from this definition that every set is a subset of itself.

Definition. Set A is a proper subset of set B iff all the members of A are also members

1 of 16 1/13/2009 6:15 PM

C}. For example. 6. size or magnitude: they have some definite number of members. then |A| = |B|.. like the odd numbers. Definition. Sets have cardinality. 5. C} = {B. then we say A B. if S = {A. if A B. 2. It follows from this definition that no set is a proper subset of itself. |S| is a number. {A.Peter Suber. but not all the members of B are members of A. and 1. This definition applies to infinite as well as to finite sets. Notation. as opposed to an ordinal number. In this brief exposition. If sets A and B can be put into one-to-one correspondence. The cardinality of set S is |S|. but it is used as an ordinal number when we say "the third pig built with brick" or "curtain number three". Technically. for example. like the even numbers. For example. that is. rather than the presence. A B. 4. C}. then we know that the nth term of one sequence will have a counterpart in the nth term of the other. However. finish it) in finite time. 3. Notation. A. Definition. or. A cardinal number answers the question how many? An ordinal number answers the question which one? Natural numbers are used both ways in different contexts. I introduce the mathematics of infinite cardinal numbers and ignore the infinite ordinals. |S| will be an infinite number.htm of B. This doesn't matter for what follows. and vice versa. Two sets can be put into one-to-one correspondence iff their members can be paired off such that each member of the first set has exactly one counterpart in the second set. Notation. and for them no such sequences can be constructed. Two sets have the same cardinality iff they can be put into one-to-one correspondence. B. most of our proofs will show the absence or failure. When S is an infinite set. We will soon see that there are infinite sets larger than the set of natural numbers (Theorem 3 below). for example. Putting two infinite sets into one-to-one correspondence is an infinite task. and each member of the second set has exactly one counterpart in the first set.%20Infinite%20Sets. Hence while S is a set. then |S|=3. But they do not have ordinality: their members are not in any particular order.. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. |S| is a cardinal number. To show that an infinite set. for cardinalities of that magnitude. and we don't pretend that we can do it (that is. we need only produce a rule-governed sequence for each set which runs through the members without omission or repetition. The cardinality of a set is the number of members it contains. If we can do so. Definition. 3 is used as a cardinal number when we say there are "three blind mice" or "three bags full". It follows from the last three definitions that set A has a larger cardinality than set B iff both (1) a proper subset of A and the whole of B can be put into one-to-one 2 of 16 1/13/2009 6:15 PM . guaranteeing one-to-one correspondence all the way out. B. can be put into one-to-one correspondence with another. but it might help you remember the term cardinality.... of one-to-one correspondence.

and (2) the whole of A cannot be put into one-to-one correspondence with any proper subset of B. but by calling it a "reminder" I'm hoping to get on your good side.. is countable and denumerable. {C}. The set of natural numbers is designated by N. We could avoid fancy new terms to prevent confusion. baseball teams.%20Infinite%20Sets. "À" is the first letter of the Hebrew alphabet. when S is finite. Now you know how many natural numbers there are: À0. Reminder..htm correspondence. . {A. The power set of a given set always contains the given set itself and the null set... {A. The natural numbers are the whole positive numbers (sometimes called the "counting numbers"). -2. {A}. including zero: 0. A set is denumerable iff its cardinality is exactly À0. The infinite set. So far we've only invented a name (numeral) for the number of natural numbers. 1. However.. Reminder. Sets with a larger cardinality than N are uncountable. As you can see from the example above. The power set of S is *S.Peter Suber.B}.C}. The null set is countable.C}. then *S = {{A. The finite set. Definition. We will assume the power set axiom. when S has 3 members.B. pronounced Aleph. is countable. For example. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. {B. A transfinite number or transfinite cardinal is the cardinality of some infinite set. In general. then |*S| = 2 | S |. 1. Definition. "infinite" has many imprecise and non-technical uses —for example. Hence À0 = |N|. Notation. N. B.. A set is countable iff its cardinality is either finite or equal to À0. The power set of a set S is the set of all the subsets of S. 3 of 16 1/13/2009 6:15 PM . the members of *S are other sets.. 3 . 2.-3. The integers are the natural numbers plus their negative counterparts. 0. "À0" is pronounced Aleph-null or Aleph- nought. Byronic poems). Notation. |*S| is the cardinality of the power set of S. C}. by definition. This is really a definition. the infinite setting on a camera's range-finder— so it often helps to use a technical term to avoid ambiguity. Definition.. -1. Note that while the members of S may be any sort of things (bugles. A set is uncountable iff its cardinality is greater than À0. The number of natural numbers is designated by À0.C}. {B}. We will justify the zero subscript when we prove that no infinite set has a smaller cardinality than the set of natural numbers (Theorem 6). that all sets have power sets. If we use the term "infinite" in a restricted and precise way. Notation. then "transfinite" is just a synonym for it. {A. B.e. then *S has 23 or 8 members. But this is not profound. C}. i. 2. 3. Ø}. if S = {A.

or |N| = |Z|. 2.. and (2) some additions. The real numbers are the rational numbers plus the irrational numbers. The set of integers has the same cardinality as the set of natural numbers. 1. 3. 1. is also false.htm Notation. Notation. We started with the natural numbers. even infinite additions. 2. both positive and negative. If each addition did increase cardinality. Reminder. For example. therefore.75 is a rational fraction because we can express it as the ratio of two integers. But if we alternate one positive and one negative integer. 4. it's wise to remember that its negation would also have violated our intuitions. and then added infinitely many irrational fractions to get the reals... then added infinitely many rational fractions to get the rationals. Proof. |N| < |Z| < |Q| < |R|.. Left in that form they would be hard to put into one-to-one correspondence with the natural numbers. 3. 5. (1) Not all infinities are equal in cardinality. do not increase cardinality. 2. -3.%20Infinite%20Sets. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber.. The set of rational numbers is designated by Q. and see how it violates our intuitions. 1. then they form a single infinite sequence. or in short: |N| < |Z| < |Q| < |R| But is this true? Here we encounter one of the first points at which the mathematics of the infinite violates our intuitions.. The rational numbers are the integers plus the rational fractions (those that can be expressed as the ratio of two integers).3. 6. 3/4. that would violate our sense that the cardinality of a set grows when we add members. and we can easily put them into one-to-one correspondence with the natural numbers. namely. -2. 4 of 16 1/13/2009 6:15 PM . 2. 7.Peter Suber. Both the intuitions mentioned in the preceding paragraph are false. For example. The irrational numbers are the fractions that are not rational numbers. At the very beginning. It's tempting to conclude that with each infinite addition we increased cardinality...) cannot be expressed as the ratio of two integers. that would violate our sense that "infinity is infinity" —or that all infinite sets are equal in cardinality.. Therefore it is an irrational number.. The set of integers is designated by Z. But if any of these additions did not increase cardinality.. The string of inequalities. we can prove that pi (3. 1. we should not demand conformity with our intuitions so much as clear definitions and rigorous proofs. Foreshadowing. 0.. -1.. 3. Notation. Therefore it is a rational number. Reminder. we note that they trail off with three dots in two directions: .14159. Theorem 1. The set of real numbers is designated by R. then added infinitely many negative whole numbers to get the integers. 0. 0. 0. If we try to list the integers. Before we learn the truth.

Proof. 4/1. every rational number will be represented infinitely many times.Peter Suber. 2/2.. The two shaded axes list all the possible natural number numerators. that between any two there are infinitely many others. For example. each of which is equal to 1. In fact.. this violates the conditions of strict one-to-one correspondence. This method of enumerating the rationals includes each rational number more than once. 4. Cantor used it often and so will we. 1/3. The interior part of the table uses the axes to compose all the rational fractions..htm This method of alternating the members of two infinite sequences in order to make a single sequence is called interlacing. 1/2. It's much harder to find a method for putting the rational numbers into one-to-one correspondence with the natural numbers than it was for the integers.. 3/3.. then diagonally down and to the left. The reason is that the rational numbers are dense.... between 123/987 and 124/987 there is 123.. It also means that for any given rational number. 1/4 2/4 3/4 4/4. for we are then 5 of 16 1/13/2009 6:15 PM . or |N| = |Q|. 2/3. and so on. which are all the rational numbers.. there is no such thing as the next greater.5/987..%20Infinite%20Sets. For example. But as the following table shows.. in fact... This means. 3 1/3 2/3 3/3 4/3. But Cantor found a very clever method for laying out the rational numbers so that they can be put into one-to-one correspondence with the naturals. 3/1. between any two of them there is a third. Naturals 0 1 2 3 4 5 6 7 8 9. it will include 1/1. Now we read off the rationals in this order: starting with the 1/1 in the upper left corner. 3/2. eventually passing through every cell in the grid. rational number. or the next lesser.... 1/4. and all the possible natural number denominators.. The set of rational numbers has the same cardinality as the set of natural numbers. 2/2. But in fact we can leave the duplicates in the enumeration without harming the proof. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. 2/1. this sequence can be put into one-to-one correspondence with the natural numbers: Rationals 1/1 2/1 1/2 1/3 2/2 3/1 4/1 3/2 2/3 1/4. that is. Technically.. we move right to 2/1. then diagonally up and to the right. Denominators 1 1/1 2/1 3/1 4/1.. The following table shows Cantor's method for putting the rationals and naturals into one-to-one correspondence: Numerators 1 2 3 4. Theorem 2. which resolves into 247/1974. The first 10 numbers in our journey are: 1/1. 2 1/2 2/2 3/2 4/2.. then down..

. i.%20Infinite%20Sets... take the "yes" written in red. Let this hypothetical pairing off be represented by the leftmost columns of the following table. For example.. In this way. and in Theorem 16 we will see that the additions to the naturals required to generate the real numbers do indeed create an uncountable set. it is uncountable.. It is in the column of natural number 1. and in the row of the set of odd numbers.. and at most as numerous as B.e.. We already know that the rationals are at least numerous as the naturals (because the naturals are a proper subset of the rationals). nos. In the shaded row of the table we again enumerate the natural numbers. move diagonally downward and to the right... for it differs from the first row in the first term. The "yeses" and "noes" in the body of the table below them tell us whether the natural numbers in the shaded row above them are members of the sets to the left of them.htm proving that the rationals plus the duplicates is still no greater in cardinality than the naturals. This means that the rationals are at most as numerous as the naturals. 0 evens yes no yes no. 1 odds no yes no yes. . From the "yes" in the upper left corner of the block of "yeses" and "noes". Conversely.. and so on.. members of correspondence the set listed at left? with the sets thereof (hypothetically) 0 1 2 3. ... .. Our next theorem shows that some sets can indeed be larger than the natural numbers.. . Theorem 3.. We give a negative proof. Therefore. The power set of the set of natural numbers has a greater cardinality than the set of natural numbers. we represent every set of natural numbers (down the left column) as an infinite string of "yeses" and "noes". . The resulting infinite string of "yeses" and "noes" is demonstrably different from every row of the infinite table. 2 squares yes yes no no. or |*N| > |N|. from the second row in the second term. the rationals are exactly as numerous as the naturals.. But we cannot generalize too quickly and say that such additions will never increase the cardinality of the original set. When 6 of 16 1/13/2009 6:15 PM . The nat... and assume that the natural numbers can be put into one-to-one correspondence with the sets of natural numbers. Toggle each "yes" we encounter to a "no" and vice versa. Proof. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. (Can you see why if some set A is at least as numerous as another set B.. following the yellow cells. it says "yes" because 1 is indeed a member of the set of odd numbers.Peter Suber.. then it must be exactly as numerous as B?) In Theorems 1 and 2 we saw that two kinds of infinite addition to an infinite set did not increase the cardinality of the original set. we can read any string of "yeses" and "noes" as code for a particular set of natural numbers. in 1-1 Are these nos.

We'll use diagonalization often. and thereby constructing a new item that differs from every row in some definite respect. then it is sad. which is S. in which we proved that the power set of a particular set. had a greater cardinality than the original.Peter Suber. N. But this contradicts our assumption that we had exhaustively listed and paired off all the sets of natural numbers. x. the most important skeptics of Cantor's results. for one of its proper subsets. This is called simply Cantor's Theorem. the set it represents is for the same reason demonstrably different from every set yet listed on the table. 3. Proof.. The present theorem is trivial for finite sets. which is the same thing as to conclude that the desired conclusion is true. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. It generalizes the previous theorem.}. The power set of the naturals must have the greater cardinality. Clearly S is one of the subsets of A.htm we read it as set of natural numbers.. This technique of moving down the diagonal of a table. The cardinality of the power set of an arbitrary set has a greater cardinality than the original arbitrary set. Again we supply a negative proof. then it is happy. finite or infinite. can be put into one-to-one correspondence with the naturals. refuse to use negative proofs in dealing with infinite sets —although they are willing to use negative proofs in other areas of mathematics. {3}. 1. which is S. is called diagonalization.%20Infinite%20Sets. Is x happy or sad? If x is happy. Let S be the set of sad members of A. Therefore S is paired off with one of the members of A. {1}. However. If x is sad. we used a negative proof. If the naturals and the sets of naturals cannot be put into one-to-one correspondence. then it is happy. say. but is fundamental for infinite sets. Our assumption implies a contradiction and is therefore false. then one has a larger cardinality than the other. 2. called intuitionists. We're almost done. 0.. In a negative proof we assume the negation of the desired conclusion and show that that assumption leads to a contradiction. then we conclude that assumptions is false. and if it is sad. and the natural numbers cannot be put into one-to-one correspondence with the sets of natural numbers. or whether we should distrust negative proofs in those domains where there is no possibility of verifying our results through intuition. otherwise it is sad. {2}. but that would make it happy. negative proofs are common in this branch of mathematics. To establish Theorem 3. Take any one of the supposed ways of pairing off the members of A with the subsets of A. but that would make it sad. So if x is happy. It is one of Cantor's most ingenious methods for gaining leverage on infinite sets. and assume that the members of A can be put into one-to-one correspondence with the subsets of A. {{0}. or |*A| > |A|. then x is a member of the set to which it is paired.. Theorem 4. Therefore there are more sets of natural numbers than natural numbers. Let A be an arbitrary set of any cardinality. Therefore our assumption is false. So the members of A cannot be put into one-to-one correspondence with the subsets 7 of 16 1/13/2009 6:15 PM . Since it is so hard to grapple with infinite sets directly. then x is not a member of the set to which it is paired. changing every row in some way. The issue between Cantor and the intuitionists is whether this refusal is arbitrary. Let us say that if a member of A is paired with a subset of A of which it happens to be a member. It is a special case of the rule of inference. by which we infer that any statement that implies a falsehood is false.. called modus tollens.

if the members of A are A1. it is the smallest transfinite cardinality. then they can be put into one-to-one correspondence with this subset of *A: {{A1}.. and another.. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. call it S1. Let S be a set of any infinite cardinality. If we follow the notation for finite sets. and so on. In this way we may carve out a denumerable proper subset from S. If S is a set of any infinite cardinality. Hence the cardinality of a denumerable set is not greater than any transfinite cardinality. For example. This follows directly from Cantor's Theorem (Theorem 4). or |*S| > |S|. No infinite set has a smaller cardinality than a denumerable set (or. {A3}. there is an infinite series of infinite cardinal numbers. or. But we make the most important of them explicit in the next theorem.." Now that we have proved there is an infinite series of infinite cardinals. Theorem 6. Proof. Speaking of the infinite series of infinite numbers which Cantor proved to exist. A2. none is smaller than the set of natural numbers). for every set. namely.." and (in 1926).. If so.}. Many profound consequences follow directly from Cantor's theorem. {S1. there is a greater set. then they cannot have the same cardinality. It follows that there is no set of all sets. S3. a clarification which alone dispells many of the ancient conundrums and paradoxes surrounding the infinite. S2. for A can be put into one-to-one correspondence with a proper subset *A. Theorem 5. without emptying S. then its power set has a greater infinite cardinality. "This appears to me to be the most admirable flower of the mathematical intellect and one of the highest achievements of purely rational human activity. "No one shall drive us from the paradise which Cantor created for us.}. S2. or no set of everything. each time without emptying S. this corollary focuses on the important consequence for infinite sets. then the larger one must be *A. it is well to prove that À0 truly designates the smallest of them. {A2}. But if A and *A cannot be put into one-to-one correspondence.. This theorem also implies that. But S had any infinite cardinality. namely. This theorem asserts that for any infinite cardinality. Clearly we may take away one of its members. there is a larger infinite cardinality. the cardinality of its power set.htm of A. David Hilbert said (in 1910) that.%20Infinite%20Sets. Hence all infinite sets have at least one denumerable proper subset. A3. and say that a set of cardinality a has a power set of cardinality 2a. We will meet some of the infinite cardinals larger than N shortly. Hence. for each transfinite cardinal a. S3. 8 of 16 1/13/2009 6:15 PM .. Cantor's theorem applies equally to finite and infinite sets. It follows from Theorem 5 that "infinity" is not synonymous with "totality". then this theorem asserts that 2a > a.Peter Suber. We may take another.

If we add element A to the set of natural numbers.. Theorem 8. 1. even though there is an infinite series of infinite cardinals (Theorem 5). which is demonstrably the smallest infinite cardinality (Theorem 6).. which one of them is À1.) Therefore. A6. {A1. then clearly the result would not be a denumerable set. If {A1. Theorem 7. and so on. In fact we may add denumerably many new members. then we can separate out the set of even-numbered members. Proof. À0 + n = À0 for every positive integer n.}. À0 designates the cardinality of the natural numbers by definition. À0 + 2 = À0.}. 3. So we cannot conclude in general that À0 . C. can be put into one-to-one correspondence with the natural numbers if we interlace its members thus: A.. À1 is by definition the next greater infinite cardinality after À0. The enlarged set.}. Proof. and so on. and so on. then there is no smallest infinite set: they nest downward ad infinitum without ever losing infinite cardinality..}. For if the denumerably many members we subtracted happened to comprise the entire membership of the original denumerable set. Note that this theorem only applies to the removal of certain denumerable subsets from a given denumerable set. As a corollary it follows that we may add a second member. Theorem 9. 9 of 16 1/13/2009 6:15 PM . then discard one of them. C. to the original denumerable set without increasing its cardinality. {A2. then the resulting set can still be put into one-to-one correspondence with the natural numbers. the other one has the same cardinality as the original. 1 with 2..%20Infinite%20Sets.. then the smallest infinite set is a denumerable set or one with a cardinality of À0. 0.. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber.. As we will see in the section on the continuum hypothesis below.Peter Suber. 1. 1. C.. it is impossible to say.. A4. But if the largeness of an infinite set is measured by the largeness of its proper subsets. each of which is denumerable. which one is À2. the result will have the same cardinality as the original. {A.. A3. B. and a third. A2. and so on. from the set of odd-numbered members. 0 with 1.. A5. This justifies the zero subscript. giving us {A.. 0. 0. or À0 + 1 = À0. We need only regard the given denumerable set as two denumerable sets interlaced... If we remove denumerably many members from an uncountably large set. then 'unlace' them. A denumerable set may have denumerably many members removed (in certain ways) without reducing the cardinality of the original set. in standard set theory. just as À2 is by definition the next greater cardinal after À1. 2. (See Theorem 1. A3. A would be paired with 0. A.À0 = À0. If the "largeness" of an infinite set is measured by its cardinality.. À0 + À0 = À0. B. If we discard one of the resulting sets. 2.. À0 + 3 = À0..htm We know from the proof of Theorem 6 that every infinite set has an infinite proper subset which consists of the original set minus denumerably many members.} is the original denumerable set. 2. A denumerable set plus a new member is still denumerable. B. This fact has a beautiful implication.

But by the present theorem. To the second. a chapter. in certain ways. there are only countably many names for anything. a book. Let us say that a text is a finitely long string of symbols. Together with the trivial truth that no finite set can be put into one-to-one correspondence with any of its proper subsets. Definition. But we know from Theorem 3 that there are uncountably many sets of natural numbers. Theorem 11. then they can be put into one-to-one correspondence. assign the numeral "100". To the first symbol. If S is countable. and its absence as the defining condition of finitude. whose symbols are chosen from a finite set (or alphabet). replace the symbols with their corresponding numerals. for example.Peter Suber. It may therefore be taken as the defining condition of infinite magnitude. Hence. Proof. This theorem has deep consequences. There are only countably many texts. this theorem establishes the important result that all and only infinite sets possess the property that they can be put into one-to-one correspondence with at least one of their proper subsets. Theorem 10. For our purposes here we needn't even say that the string conforms to the rules of some grammar. a sentence. If we do so. Concatenate the numerals so that. a paragraph. But there are only À0 natural numbers. say D. To the third. Now in each text. This theorem also means that there are at most countably many theorems in any axiom system (in which theorems are finitely long). Proof. List them exhaustively in any order. In general. Or conversely (by Theorem 9). uncountably many real numbers cannot be named. We will soon prove (Theorem 16) that there are uncountably many real numbers. to the nth symbol in the alphabet we assign the numeral in which "1" is followed by n zeroes. or a library. Therefore there are at most À0 texts. assign "1000" and so on.htm Proof. "10 100 1000 10" becomes "10100100010". then Theorem 8 tells us that we may produce a proper subset of S without reducing S's cardinality. By definition there are only finitely many symbols in the "alphabet" from which we may compose texts. When interpreted. the text may be a name. without reducing the cardinality of D.%20Infinite%20Sets. then we thereby remove denumerably many members from U without reducing the cardinality of U. And we know from Theorem 8 that we can remove denumerably many members from D. It follows that this property is a necessary and sufficient condition for being an infinite set. Every infinite set can be be put into one-to-one correspondence with at least one of its proper subsets. We know from Theorem 6 that U has at least one denumerable proper subset. only a countable subset of the real numbers can be named. then we have the same result using Theorem 9 instead of Theorem 8. Let U be any uncountably large set. But if S and that proper subset have the same cardinality. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. We now have a method for converting every distinct text into a distinct natural number. If there is at least one truth of 10 of 16 1/13/2009 6:15 PM . If S is uncountable. assign the numeral "10". Let S be a set of any infinite cardinality.

to the left of C. the numerical continuum has the same cardinality as the linear continuum. and the line we draw will intersect AB somewhere. 11 of 16 1/13/2009 6:15 PM . It follows that every axiom system intended to capture number theory is doomed to incompleteness.g. and the line we draw will intersect AD somewhere. and a new point. Definition. Label the point at the upper left. Therefore. we may add back the point D. with a new point. From point C we can draw a line through any point of AD. there are uncountably many more truths of number theory than there are texts. that some number n belongs to that set). We have yet to say how many points are on AD (minus D itself) or the infinite ray AB. which we omitted above. the number of points in a finite segment equals the number of points on an infinite line. D'. D. Notation. though still finite segment. minus D itself. Hence c = |R| by definition. B'. the cardinality of the set of real numbers. in this way. Until I can draw a GIF for this theorem. and those on AB can be put into one-to-one correspondence. except D itself. we can pair any point on AB with exactly one point on AD. The number of points on a finite line segment is the same as the number of points on an infinite ray. Now since any infinite cardinal plus one equals the original infinite cardinal (Theorem 7).Peter Suber. Therefore. But the two line segments together make one longer. Proof. But this means the points on AD. from point C we can draw a line through any point of the ray AB. Imagine the mirror image of the figure used in Theorem 12 (a "backward Z"). that at the lower left A. C. then there will be uncountably many truths of number theory. to the left of A. The proof is a simple variation on the proof for Theorem 12. and that point B is somewhere along it to the right of A. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. without changing the cardinality of the set of points on the segment. Theorem 13. in this way we can pair any point on AD with exactly one point on AB. Theorem 12. Imagine that the bottom bar continues infinitely to the right. imagine a large letter "Z". Let c (lower-case "c" for "continuum") designate the cardinality of the continuum —or equivalently. Let segment CD be parallel to ray AB. Hence they contain the same number of points. or in the jargon. The number of points on a finite line segment is the same as the number of points on an infinite line.%20Infinite%20Sets. We would then prove that two line segments (AD and AD') together contain the same number of points as the infinite line. that at the upper right. The number of real numbers is the same as the number of points on an infinite line.htm number theory for each set of natural numbers (e. but we know it will be some infinite cardinality. Conversely.

. Together they make a longer. Theorem 16. this theorem asserts that c > À0. . Proof.. The power set of the set of natural numbers has the same cardinality as the set of real numbers. Therefore the assumption is false and the reals cannot be put into one-to-one correspondence with the naturals.. hence one which has c points. the cardinality of the set of real numbers is greater than the cardinality of the set of natural numbers. Therefore. though still finite line segment. 0. Now we have only to prove that the number of reals between 0 and 1 is the same as the number of all the reals. 2 0. Hence the whole line has as many points. move diagonally down and to the right. But from Theorem 13 we know that each segment has c points. dddd. from the second in the second place. for example. dddd. But this contradicts our assumption that we have exhaustively listed all the real numbers between 0 and 1. . c + c = c. À0. such segments. or À0. Clearly the segments could be put into one-to-one correspondence with the natural numbers. A second of arbitrary length has c points. 3c = c. and nc = c for every positive integer n. like a ruler. as the number of segments.5 becomes 0. times the number of points per segment.%20Infinite%20Sets. Therefore we have as corollaries.. Express terminating decimal fractions as non-terminating decimal fractions. Now starting with the upper left d. or |*N| = |R| = c.4999. One finite line segment of arbitrary length has c points (Theorems 12 and 13)... The set of real numbers is uncountable. In this way we construct a new decimal fraction that is demonstrably nowhere on our list.. Now under our assumption pair off the natural numbers with the reals between 0 and 1. Take the set of real numbers between 0 and 1. dddd.Peter Suber. Hence there are denumerably many. and changing 9 to 0. But the whole line also has c points (by definition of c). Think of an infinite line as marked off into unit segments. c. or |R| > À0. for it differs from the first one in the first decimal place. following the yellow highlights. incrementing it by one. where d is some digit. Theorem 17. Theorem 15. But the reals are at least as numerous as the naturals and cannot have the smaller cardinality. Proof. Again we use a negative proof. for example.. À0 — c = c. thus: 0 0. and so on. c. Proof.htm Theorem 14... 2c = c... "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. Equivalently. But this follows directly from Theorem 13. Change each d we encounter to some other digit.. 12 of 16 1/13/2009 6:15 PM . Assume that the real numbers can be put into one-to-one correspondence with the natural numbers. 1 0.

Each of these new. rather than a pair of real numbers.1323530303030.. can be put into one-to-one correspondence with the denumerable strings of "1's" and "0's". We know from the proof of Theorem 3 above that every set of natural numbers can be represented as a denumerable string of "yeses" and "noes". just tack a decimal point (or a zero and then a decimal point) onto the left end. we may use the notation introduced in Theorem 5 to restate this theorem thus: c = 2À0 (2 to the power of À0. c). But there are at most countably many such finite strings. Now interlace these two numbers as we did in Theorems 1 and 7. then interlace these two decimal fractions digit by digit to make 0.. for example... to correspond to each point inside the square. Theorem 18. "I see it...0111.htm Proof...111000. The set of points in a square has the same cardinality as the set of points on one of its edges (namely.. as we saw in Theorem 16. if you don't believe me. think of left and bottom edges of the square as collinear with the x and y axes of a Cartesian coordinate system. and cut off the trailing zeroes." at some point in their expansions.. therefore.. Therefore. 111000.1000. 0. 0.Peter Suber. and regard the result as a fraction in base two. in case your browser choked on this expression). For example.. Intensively for the three years from 1871 to 1874 Cantor labored to prove this theorem false. (For clarity. see the post-script below. we will produce distinct. can be put into one-to-one correspondence with the real numbers between 0 and 1 if we put a decimal point at the left end each denumerable string of 1's and 0's. and the y axis position is 0. c — c = c. (We can purge this set of numeral strings of repetitions without reducing the cardinality of the set. if the x axis position is 0.. For example..%20Infinite%20Sets. Post-script. But since they correspond to the points inside a square we also know that there are at least c of them. = 0. 13 of 16 1/13/2009 6:15 PM . in turn. Every distinct point inside the square. First. according to Theorem 11. the other its position on the y axis.333. there are exactly c of them.. interlaced numbers is a real number between 0 and 1. These denumerable strings. these real numbers between 0 and 1 can be put into one-to-one correspondence with all the reals. These strings. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. We know from Theorems 12 and 13 that the number of points in a finite line segment is c. We know this because if we take all the fractions that consist of "000.. Remembering that |N| = À0 (by definition)... one is its position on the x axis.4999. has a distinct pair of real numbers to identify it. Then he surprised himself by proving it true.. To convert a string of 1's and 0's into a binary fraction.. finite strings of 1's and 0's..) There are at most countably many such repetitions. We now have a single real number..." he wrote to Dedekind. which equals 1/2 + 1/4 + 1/8 + 0/16 + 0/32. Hence to delete the countably many repetitions would leave us with the same uncountable cardinality we started with.. Some of these fractions will equal others..) Finally. as we saw in Theorem 13. in turn. Now we know from Theorem 16 that there are uncountably many real numbers between 0 and 1.. remember that in base 10.. We know from Theorem 16 that there are at most c of these. = 0. Proof. Now we need only show that the number of points inside a square is also c.125000. according to Theorem 9. Theorem 19. but I don't believe it..5000. becomes 0.

. Hence the cells and the natural numbers can be put into one-to-one correspondence. The set of all points in À0-dimensional space has the same cardinality as the set of all points in a finite line segment. In fact. The set of all points in a cube has the same cardinality as the set of all points in one of its edges. Yet each cell contains a natural number. Pick one cell arbitrarily. but perhaps you can visualize it. 3-dimensional. The rest of the proof is the same. But the plane is À0 cells 'high'. Go to the cell above it and number that cell 1. Each point inside the cube will have three coordinates. — c = c. Proof. it is equivalent to one of the corollaries of Theorem 19. or À0.Peter Suber. Theorem 22. and À0 cells 'wide'. namely. Proof. Our spiraling path which puts the natural numbers into one-to-one correspondence with the cubes will be more complex. First we show that there will be denumerably many. and number it 0. then the set of all points 14 of 16 1/13/2009 6:15 PM . not two. the number of points in the infinite plane is the number of cells. But another way to ascertain the number of such cells is to take the product of the 'length' of the plane (in cells) and the 'width' of the plane (in cells). The set of all points in infinite. such square cells. namely. Theorem 20. or À0. times the number of points in a cell. like graph paper. c. Theorem 24. Second we note that each cell contains c points. The result will be a spiral that would eventually cover the plane. c. As corollaries of this theorem. The set of all points in an infinite plane has the same cardinality as the set of all points in a finite line segment. The proof is a simple variation of that for Theorem 18. Think of the plane as marked off into an infinite number of square cells. namely. or in short. Continue in this way to circle the "0" cell. cn = c when n is any positive integer. The rest of the proof is the same as for Theorem 18. we have c — c — c = c. The proof is a variation on that of Theorem 21. we already know the number of cells (from Theorem 21) to equal À0. c. not two. except that we use Theorem 20 rather than Theorem 18. Theorem 21.htm This theorem follows directly from Theorem 18. Go one cell to the right and number it 2.. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. À0. Hence we interlace three decimal fractions. which we know is equal to c (by Theorem 15). Therefore. under Theorem 18. cells in the infinite graph paper of the Euclidean plane. rather than the plane into squares. c (by Theorem 18). and c — c — c — . namely. À0 — À0 = À0. If the set of all points in 3-dimensional space is c3.%20Infinite%20Sets. However. Theorem 23. c). Hence the number of cells is À0 — À0. The proof of this theorem was already encapsulated in the proof of Theorem 21 when we showed that there were denumerably many. Euclidean space has the same cardinality as the set of all points in a finite line segment. We divide the space into cubes.

That is the main reason why most mathematicians accept CH today. After Euclid's parallel postulate. Similarly. CH was the first major conjecture to be proved undecidable by standard mathematics. consumed most of Kurt Gödel's 36 years at the Institute for Advanced Study. c = |*N| = |R|. Since we already know (from Theorem 17) that 2À0 = c. we can march through the transfinite cardinals with the power set operation. And this in turn entailed that CH could be assumed and used in standard set theory without introducing any inconsistencies that were not already there. It follows that for every infinite 15 of 16 1/13/2009 6:15 PM . and rationals. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber. GCH yields the ordinary CH. Hence it also follows that c = À1. the closest thing we have to "standard" set theory. This meant that the negation of CH (notation: ~CH) could not be derived from the ZF axioms. integers. or to prove it undecidable. which in turn equals 2(À0 — À0). which (by Theorem 22) equals 2À0. But he did produce an important partial proof. Its independence also means that ~CH is just as consistent with standard set theory as CH itself. GCH implies that.%20Infinite%20Sets. and we would know the cardinality of the naturals. The continuum hypothesis (CH) asserts that there is no cardinal number a such that À0 < a < c. just as geometers can develop geometries with or without Euclid's parallel postulate. Among other things. The widespread acceptance of the assertion by mathematicians therefore attests to the general acceptance of CH in the profession. Gödel and Cohen's results show that CH is independent of the ZF axioms: neither it nor its negation can be derived from them. To prove or disprove CH was the first problem on David Hilbert's famous 1900 list of important unsolved problems in mathematics. Since by Theorem 17.Peter Suber. His failure obsessed him and caused recurring bouts of serious depression. starting with À0. and will not skip any. Definition.) Cantor formulated CH. Gödel in effect proved that CH was as harmless as the less exotic propositions that already comprised set theory. From CH it follows that the next largest transfinite cardinal after À0 is c. attempts to prove or disprove CH. But under Theorem 17. this means that CH is undecidable in ZF. and spent the last years of his career unsuccessfully trying to prove it. The generalized continuum hypothesis (GCH) asserts that 2Àa = Àa+1 for every positive integer a. This has allowed set theorists to develop set theories with and without it. When a=0. Gödel showed that CH cannot be disproved from the axioms of Zermelo-Fraenkel (ZF) set theory. Definition.htm in À0-dimensional space is cÀ0. (Without CH we have essentially no idea which Aleph corresponds to c. Set theory with ~CH rather than CH is usually called Non-Cantorian set theory. In 1963 Paul Cohen showed that CH cannot be proved from the ZF axioms. The commonly heard assertion that c = À1 assumes CH. which (by Theorem 17) equals c. the latter number equals (2À0)À0. In 1938. but not the cardinality of the reals. we could use any of the latter two expressions in place of c in stating CH. Together. CH allows us to say that 2À0 = À1.

. . 16 of 16 1/13/2009 6:15 PM . Metalogic. |**N|. "Infinite Sets" file:///I:/Math%20Misc/Peter%20Suber.A. North-Holland Pub. B. Some of the logic symbols in this file are GIFs. Abraham Fraenkel.. University of California Press. Some are HTML characters using the Symbol Font. then |*S| = Àn+1. Department of Philosophy. 1955.Peter Suber. Peter Suber. 1953.%20Infinite%20Sets. Co. Richmond. Jourdain as Contributions to the Founding of the Theory of Transfinite Numbers. see Alan Wood's guide to its symbols. My exposition of Cantor's results is indebted to three more recent authors: Stephen Cole Kleene. He published two summary statements of his results in 1895 and 1897. Introduction to Metamathematics. peters@earlham. 1971.. Copyright © 1998. Peter Suber. Most of the theorems and proofs in this crash course were discovered by Georg Cantor (1845-1918) and published in a series of monographs starting in 1870. U. 1952. |*N|. 47374.S. Co.htm set S.. Abstract Set Theory. Bibliographic note. and Geoffrey Hunter. North-Holland Pub. which have been translated into English by Philip E. if |S| = Àn.edu. Indiana. or that the only transfinite cardinals are |N|. Dover Publications. |***N|. see my Notes on Logic Notation on the Web.. Earlham College.