TheoremComplete

The Well-Ordering Theorem for Ordinals

Every set of ordinals is well-ordered, and any two well-ordered sets are comparable: one is isomorphic to an initial segment of the other. This establishes ordinals as the canonical representatives of well-order types.


Statement

Theorem4.5Comparability of well-orderings

Let (A,A)(A, \leq_A) and (B,B)(B, \leq_B) be well-ordered sets. Then exactly one of the following holds:

  1. (A,A)(A, \leq_A) is order-isomorphic to (B,B)(B, \leq_B).
  2. (A,A)(A, \leq_A) is order-isomorphic to an initial segment of (B,B)(B, \leq_B).
  3. (B,B)(B, \leq_B) is order-isomorphic to an initial segment of (A,A)(A, \leq_A).

Moreover, any order-isomorphism between well-ordered sets is unique.


Proof

Proof

Step 1: Uniqueness of isomorphisms. Let f,g:(A,A)(B,B)f, g: (A, \leq_A) \to (B, \leq_B) be order-isomorphisms. By transfinite induction on AA, we show f=gf = g. If aa is the least element of AA, both f(a)f(a) and g(a)g(a) must be the least element of BB. If f(a)=g(a)f(a') = g(a') for all a<aa' < a, then f(a)f(a) and g(a)g(a) must both be the least element of B{f(a):a<a}B \setminus \{f(a') : a' < a\}, so f(a)=g(a)f(a) = g(a).

Step 2: No well-ordered set is isomorphic to a proper initial segment of itself. Suppose f:AAb={aA:a<b}f: A \to A_b = \{a \in A : a < b\} is an order-isomorphism. Then f(b)<bf(b) < b. Define c=f(b)c = f(b), then f(c)<f(b)=cf(c) < f(b) = c (since ff preserves order and c<bc < b). Iterating: b>f(b)>f2(b)>b > f(b) > f^2(b) > \cdots, an infinite strictly decreasing sequence, contradicting well-ordering.

Step 3: Comparability. Define a partial isomorphism ff as follows: by transfinite recursion on AA, set f(a)=f(a) = the least element of Bf[{a:a<a}]B \setminus f[\{a' : a' < a\}], if such an element exists. The domain of ff is an initial segment of AA, and the range is an initial segment of BB.

If dom(f)=A\mathrm{dom}(f) = A and ran(f)=B\mathrm{ran}(f) = B: Case 1. If dom(f)=A\mathrm{dom}(f) = A and ran(f)=Bb\mathrm{ran}(f) = B_b for some bb: Case 2. If dom(f)=Aa\mathrm{dom}(f) = A_a and ran(f)=B\mathrm{ran}(f) = B: Case 3. If dom(f)=Aa\mathrm{dom}(f) = A_a and ran(f)=Bb\mathrm{ran}(f) = B_b: impossible, since we could extend ff by f(a)=bf(a) = b. \blacksquare


Consequences

Theorem4.6Every well-ordered set is isomorphic to a unique ordinal

For every well-ordered set (A,)(A, \leq), there exists a unique ordinal α\alpha and a unique order-isomorphism f:(A,)(α,)f: (A, \leq) \to (\alpha, \in). The ordinal α\alpha is called the order type of (A,)(A, \leq), written otype(A,)=α\mathrm{otype}(A, \leq) = \alpha.

ExampleOrder types
  • otype({3,7,15},)=3\mathrm{otype}(\{3, 7, 15\}, \leq) = 3 (any finite set of size nn has order type nn).
  • otype(N,)=ω\mathrm{otype}(\mathbb{N}, \leq) = \omega.
  • otype({0,1,2,}{}\mathrm{otype}(\{0, 1, 2, \ldots\} \cup \{*\}, where n<n < * for all n)=ω+1n) = \omega + 1.
  • otype(N×{0,1}\mathrm{otype}(\mathbb{N} \times \{0, 1\}, lex order)=ω2) = \omega \cdot 2.
RemarkThe Burali-Forti paradox

The class of all ordinals Ord\mathbf{Ord} is not a set. If it were, it would itself be a transitive set well-ordered by \in, hence an ordinal Ω\Omega. But then ΩΩ\Omega \in \Omega, which contradicts the axiom of foundation (and well-ordering). This is the Burali-Forti paradox, the first of the set-theoretic paradoxes involving "too large" collections.