11

Maxwell Rosenlicht claims in "Introduction to analysis" that a set is infinite if and only if it may be placed into one-to-one correspondence with a proper subset of itself.

He says this is self-evident because a finite set cannot be placed into a one-to-one correspondence with a proper subset of itself (because it has fewer elements), and whilst this is reasonable - I cannot follow Rosenlicht in that "the above therefore follows obviously". Why must a set be infinite just because of some property of finite sets?

  • 7
    Because infinite means "not finite". – vadim123 Oct 13 '13 at 00:57
  • 6
    @vadim123 Yes, but just because a finite set is incapable of some property, how does that mean that a set that is not finite set MUST have this property? Could it just not be (at this level of the discussion, there are of course cantor sets later etc.) that both finite and infinite sets cannot be placed into a 1-1 onto correspondence with a smaller set? – Frederick G Oct 13 '13 at 01:04

8 Answers8

8

In standard terminology, a "finite" set means one whose cardinality is a natural number, or in other words a set what is in bijective correspondence with $\{i\in\mathbb N\mid i<n\}$ for some $n\in \mathbb N$.

A set that is not in bijective correspondence with any proper subset of itself is called Dedekind-finite.

As you note, it is obvious that a finite set is also Dedekind-finite. But you're right that it is not obvious that every Dedekind-finite set is finite. In fact this is not necessarily true if we're working in a set theory without the Axiom of Choice.

If we do have the Axiom of Choice, however, we can prove easily that every set is either finite or contains a subset (not necessarily proper) that is in bijective correspondence with $\mathbb N$. In the latter case we can prove using the Hilbert's-Hotel construction that the set is not Dedekind-finite.

(Proof sketch. Let $A$ be a set and assume $A$ is not finite. Fix a choice function on the set of nonempty subsets of $A$. Construct by induction a function $f:\mathbb N\to A$ such that $f(n)$ is the chosen element of $A_n = A\setminus f(\{0,1,2,\ldots,n-1\})$. Because $A$ is not finite, $A_n$ is never empty. Then $f$ is a bijection between $\mathbb N$ and a subset of $A$.)

  • 1
    The exact quote is: "It is easy to show that a set X is finite if and only if it may be put into 1-1 correspondence with a proper subset of itself. To do this, note first that if X is finite then any proper subset has a smaller number of elements, whereas two finite sets in 1-1 correspondence must have the same number of elements. This proves the 'if' part."

    Rosenlicht defines finite sets as those in 1-1 correspondence with {1..n} for some n in the natural numbers

    I still dont get why this property of the finite sets imbues a certain property of the infinite sets.

    – Frederick G Oct 13 '13 at 01:22
  • 1
    @FrederickG: The first sentence you quote there is missing a "not" somewhere. The rest of the quote clearly purports only to prove one direction of "if and only if", namely that a finite set is Dedekind-finite. The other direction is less obvious, and requires more subtle assumptions, as I describe in my answer. – hmakholm left over Monica Oct 13 '13 at 01:26
  • 1
    You are correct, he writes "that a set X if infinite", not 'finite'! – Frederick G Oct 13 '13 at 01:28
  • @FrederickG: In any case, your answer is: "Yes, there's something missing there". – hmakholm left over Monica Oct 13 '13 at 01:30
  • I have it. I was wondering whether it is NECESSARY that every infinite set has this property, but I see that his statement precisely requires this (without delving into the issues of whether this is possible). Thank you I have it! – Frederick G Oct 13 '13 at 01:31
  • The above proof sketch is same as ProofWiki proof 2 (If $...$ fails to be rendered, maybe it is because they are not in one line). Then we can follow reference link of marty cohen's answer wikipedia proof to construct one $B$ to have "bijective correspondence" between the "proper subset" and $A$. – An5Drama May 01 '24 at 12:59
6

This is a relatively subjective issue -- we're talking about what it means, exactly, for a set to be infinite.

That said, if you agree that a finite set is, by definition, a set that cannot be put into 1-1 correspondence with itself and a set that's not finite is infinite, you get

$$ \left(A \mbox{ finite} \iff A \mbox{ cannot be put into 1-1 correspondence with a proper subset}\right) \implies \left(\lnot(A \mbox{ finite}) \iff \lnot(A \mbox{ cannot be put into 1-1 correspondence with a proper subset})\right) \implies \left(A \mbox{ infinite} \iff A \mbox{ can be put into 1-1 correspondence with a proper subset}\right) $$

alecbz
  • 1,300
  • Interesting. Yes, a good path to choose. The thing I cannot understand is how that it should be 'obvious' that your second line is correct: i.e. is it obvious that an infinite set A can be placed into a 1-1 correspondence with a proper subset of itself? – Frederick G Oct 13 '13 at 01:08
  • The second line in my "proof" is true because of contrapositives (ie, $(A \implies B) \iff (\lnot B \implies \lnot A)$). It is "obvious" that an infinite set can be put into 1-1 correspondence with a proper subset if we choose to take this as the definition of "infinite". – alecbz Oct 13 '13 at 01:26
  • Great, I have it! I feel stupid now :/ – Frederick G Oct 13 '13 at 01:36
  • This is not the standard definition. – Andrés E. Caicedo Oct 13 '13 at 01:40
  • @AndresCaicedo Is dedekind-infinite not the standard definition? – Frederick G Oct 13 '13 at 01:42
  • 4
    No. The qualifier in "Dedekind-infinite" indicates that this is something special. What we call infinite, without any adjectives, is defined as "not finite", and "finite" is defined as "in bijection with ${i\in\mathbb N\mid i<n}$ for some $n\in\mathbb N$". The argument showing that infinite sets are Dedekind-infinite uses a version of the axiom of choice. It is consistent with the axioms of set theory without choice that there are infinite sets that are Dedekind-finite. – Andrés E. Caicedo Oct 13 '13 at 01:55
  • Andrés E. Caicedo may mean the definition of "finite set" is not standard. See https://math.stackexchange.com/a/656929/1059606 and dfeuer's answer. – An5Drama May 02 '24 at 01:46
4

The given definition is usually taken as the definition of "Dedekind finite", whereas "finite" is usually taken to mean "equivalent to a natural number/finite ordinal". A finite set is always Dedekind-finite. However, in set theory without the axiom of choice, it's consistent for an infinite set to be Dedekind-finite!

Here's a proof, using the axiom of countable choice, that every infinite set has a countably infinite subset, allowing Hilbert's hotel to finish the argument.

dfeuer
  • 9,369
4

This is known as the Dedekind definition of a set being infinite. Here is more: http://en.wikipedia.org/wiki/Dedekind-infinite_set

As an exercise, you might try to show that this is equivalent to the definition stating that the set, or some subset of it, can be placed into a 1-1 correspondence with the positive integers.

marty cohen
  • 110,450
  • Thank you for the link, I understand this particular theorem now but the wikipedia link is new to me - will be learning Dedekind-finite immediately :) – Frederick G Oct 13 '13 at 01:32
4

I agree with you. The author gave a reason why a finite set cannot have a one-to-one correspondence with a proper subset of itself, but he did not give a reason why a set that does not have a one-to-one correspondence with a proper subset of itself must necessarily be finite. Therefore the only thing that he can logically assert is that if a set has a one-to-one correspondence with a proper subset of itself, it must be infinite. He gives no justification for the if-and-only-if relationship.

In other words, the author asserts that

$$(A \implies \lnot B) \implies (\lnot A \iff B)$$

Which is not sound. Having said that, the author's statement is true, despite the fact that his logic does not back it up.

nispio
  • 462
2

This answer might be a bit late but here we go.

Statement 1: A set is either finite or infinite. (Law of excluded middle)

Statement 2: A finite set cannot be put into one-to-one correspondence with a proper subset of itself. Proof.

Statement 3: So, by statement 1, if there exists a set which can be put into one-to-one correspondence with a proper subset of itself, then it should be an infinite set.

See this link for more descriptive proof!

slhulk
  • 293
2

If set $S$ is finite then there is no such proper subset. This means that if there is such proper subset for $S$ then $S$ is infinite. It's just $(A \implies B) \iff (\neg B \implies \neg A)$ for $A = [\text{$S$ is finite}]$ and $B = [\text{no such proper subset exists}]$.

Smylic
  • 8,098
1

Theorem For every set $A$, the following conditions are equivalent (assuming AC$_{\omega}$ only for $\text{(a)}\Longrightarrow\text{(b)}$):

$\text{(a)}\;A$ is an infinite set

$\text{(b)}\;A$ has a countably infinite subset

$\text{(c)}\;A$ has a proper subset to which it is equinumerous

We will prove, first, the chain of implications $\text{(a)}\Longrightarrow\text{(b)}\Longrightarrow\text{(c)}$, to then prove the chain $\text{(c)}\Longrightarrow\text{(b)}\Longrightarrow\text{(a)}$.

$\text{(a)}\Longrightarrow\text{(b)}$: Let $A$ be an infinite set. For each $n\in\omega$, let $A_n$ be the set of injective functions from $n$ to $A$. For each $n\in\omega$, the set $A_n$ is not empty, since $A$ is an infinite set. By AC$_{\omega}$, for the countable set $\{A_n|\;n\in\omega\}$, the cartesian product $\prod_{n\in\omega}A_n$ is not empty, and there exists a function $h$ of domain $\omega$ such that for each $n\in\omega$, $h(n)=h_n$ is an injective function from $n$ to $A$.

Then, we can form a $\omega$-sequence of distinct elements of $A$, first putting the value of $h_1$, then the values of $h_2$, etc.:

$$h_1(0);h_2(0),h_2(1);h_3(0),h_3(1),h_3(2);h_4(0),h_4(1),h_4(2),h_4(3);\dots$$

And then deleting each occurence of any element $a$ of $A$ after appearing for the first time. This sequence is actually infinite, because for any $n\in\omega,\;h_n$ has $n$ distinct terms, and we have deleted at most $n-1$ of them, if $n\geq 1$.

The set of all the terms of this sequence is a countably infinite subset of $A$.

$\text{(b)}\Longrightarrow\text{(c)}$: Let $C$ be a countably infinite subset of $A$. Let $f$ be a bijective function from $\omega$ onto $C$. Consider the function $g:A\longrightarrow A$ defined by: for each $a\in A$,

$$g(a)=\begin{cases} f(n+1)\quad\text{if }a\in C\text{ and }a=f(n)\text{ for some }n\in\omega\\ a\qquad\qquad\,\text{if }a\in A\setminus C \end{cases}$$

The function $g$ s injective from $A$ to $A$, so $A\approx\text{Im}(g)$, and since $f(0)\not\in\text{Im}(g),\;\text{Im}(g)$ is a proper subset of $A$ to which it is equinumerous

$\text{(c)}\Longrightarrow\text{(b)}$: Let $B$ a proper subset of $A$ such that $A\approx B$, and let $f$ be a bijective function from $A$ ontro $B$. Since $B$ is proper, let $a_0$ be an element of $A\setminus B$. Let $g:\omega\longrightarrow A$ be the sequence defined by

$$\begin{cases} g(0)=a_0\\ g(n+1)=f(g(n)) \end{cases}$$

We will prove that $g$ is injective, and therefore $\text{Im}(g)$ will be a countably infinite subset of $A$.

Suppose that $g$ is not injective. Then there exists two natural numbers $m,n$ with $m<n$ wuch that $g(m)=g(n)$. Therefore, the set

$\mathcal{N}=\{n\in\omega|\;\text{ there exists }m>n\text{ such that }g(m)=g(n)\}$

is a nonempty subset of $\omega$. Let $n_0$ be its first element.

  • In first place, $n_0\not=0$, because $g(0)=a_0,\;a_0\not\in B$, and if $n>0,\;g(n)\in\text{Im}(f)=B$
  • Since $n_0>0$, we have that if $m\in\mathcal{N}$ is such that $m>n_0$ and $g(n_0)=g(m)$, then

$$g(n_0)=f(g(n_0-1))=f(g(m-1))=g(m)$$

And, since $f$ is injective

$$g(n_0-1)=g(m-1)$$

Which contradicts the choice of $n_0$ as the first element of $\mathcal{N}$

$\text{(b)}\Longrightarrow\text{(a)}:$ Suppose that $A$ has a countably infinite subset $B$, and let $f:\omega\longrightarrow B$ be a bijection. If $A$ was finite, we would have that $A\approx n$ for some $n\in\omega$, for some bijective function $g:A\longrightarrow n$. The composition $g\circ f:\omega\rightarrowtail n$ would be injective, and its restriction to $n+1$ an injective function from $n+1$ to $n$, in contrary to the pidgeonhole principle.

Akerbeltz
  • 2,853
  • IMHO, 1. here $n$ may mean 'von Neumann ordinal' by the form of '$\omega$-sequence'. 2. Maybe one typo $h(n)=A_n$. – An5Drama May 01 '24 at 12:55
  • The 2nd proof may mean $g:A\longrightarrow A\setminus{f(0)}$ and $g$ is bijective based on another answer https://math.stackexchange.com/a/3255606/1059606 of yours. Then your first 2 proofs share the similar ideas as the wikipedia proof https://en.wikipedia.org/wiki/Dedekind-infinite_set#Proof_of_equivalence_to_infinity,assuming_axiom_of_countable_choice. The 3rd proof shares the similar idea as https://proofwiki.org/wiki/Equivalent_Conditions_for_Dedekind-Infinite_Set#(1)_implies(2). You may mean $n_0\in \mathcal{N}$ by your definition. – An5Drama May 01 '24 at 13:46
  • The 4th proof means $|w|>|n|$ so it is impossible for $g\circ f$ to be injective. Thanks for your sharing. – An5Drama May 01 '24 at 13:47