Exercise 6.5.16

(Characterization of Ordinal Exponentiation) Let α and β be ordinals. For f : β α , let s ( f ) = { ξ < β f ( ξ ) 0 } . Let S ( β , α ) = { f f : β α  and  s ( f )  is finite } . Define on S ( β , α ) as follows: f g if and only if there is ξ 0 < β such that f ( ξ 0 ) < g ( ξ 0 ) and f ( ξ ) = g ( ξ ) for all ξ > ξ 0 . Show that ( S ( β , α ) , ) is isomorphic to ( α β , < ) .

Answers

First we define summation notation for ordinals. Since ordinal addition is not commutative we define summation notation such that each additional term is pre-added to the previous terms, i.e. added on the left. So, for example, we have

n = 1 5 α n = α 5 + α 4 + α 3 + α 2 + α 1 .

We also adopt the convention that

k = n m α k = 0

any time n > m .

Proof. First we note that if β = 0 = then the only function from β to α (it could even be here that α = 0 = ) is the vacuous function . Hence S ( β , α ) = { } , which is clearly vacuously isomorphic to (in fact identical to) α β = α 0 = 1 = { 0 } = { } . Thus in the following we assume that β 0 , which implies that α 0 as well since functions from β to α cannot exist when β 0 = but α = 0 = .

Also if α = 1 = { 0 } (and still β 0 ) then there is clearly only a single function f : β α , namely that where f ( ξ ) = 0 for all ξ β . Hence S ( β , α ) = { f } , which is clearly trivially isomorphic to α β = 1 β = 1 = { 0 } . Thus in what follows we shall assume the more interesting case when α > 1 (and β > 0 ).

Now we define a function h : S ( β , α ) α β . For any f S ( β , α ) since s ( f ) is a finite set of ordinals it follows that it is isomorphic to some natural number n . Thus its elements can be expressed as a strictly increasing sequence { s k } for k < n where each s k β since s ( f ) β . We now set

h ( f ) = k = 0 n 1 α s k f ( s k ) ,

noting that it could be the case that n = 0 so that h ( f ) = 0 , consistent with our convention.

First we show that h ( f ) α β so that the range of h is in fact a subset of α β . We begin by showing by induction on m that

k = 0 m α s k f ( s k ) < α s m + 1 .

First for m = 0 we have

k = 0 m α s k f ( s k ) = k = 0 0 α s k f ( s k ) = α s 0 f ( s 0 ) < α s 0 α (by Exercise 6.5.7a since  α 0  and  f ( s 0 ) < α ) = α s 0 + 1 = α s m + 1 (by Definition 6.5.9b)

Now suppose that

k = 0 m α s k f ( s k ) < α s m + 1 .

It follows from Exercise 6.5.14b that

α s m + 1 α s m + 1

since { s k } is a strictly increasing sequence so that s m + 1 s m + 1 . We then have

k = 0 m + 1 α s k f ( s k ) = α s m + 1 f ( s m + 1 ) + k = 0 m α s k f ( s k ) < α s m + 1 f ( s m + 1 ) + α s m + 1 (by the induction hypothesis and Lemma 6.5.4a) α s m + 1 f ( s m + 1 ) + α s m + 1 (follows from Lemma 6.5.4 and the above) = α s m + 1 [ f ( s m + 1 ) + 1 ] (by Definition 6.5.6b) = α s m + 1 α (by Exercise 6.5.7 since  f ( s m + 1 ) + 1 α ) = α s m + 1 + 1 , (by Definition 6.5.9b)

which completes the induction. Hence we have

h ( f ) = k = 0 n 1 α s k f ( s k ) < α s n 1 + 1 α β

by Exercise 6.5.14b since s n 1 + 1 β since s n 1 < β . Clearly then h ( f ) α β by definition of ordinal ordering.

Now we show that h is an increasing function and therefore also injective. So consider any f and g in S ( β , α ) such that f g . Then by the definition of there is a ξ 0 < β such that f ( ξ 0 ) < g ( ξ 0 ) and f ( ξ ) = g ( ξ ) for all ξ 0 < ξ < β . Now let S = s ( f ) s ( g ) , which is clearly a finite set of ordinals since s ( f ) and s ( g ) are. Hence it is also isomorphic to a natural number n so that it can be expressed as a strictly increasing sequence { s k } for k < n . Moreover

h ( f ) = k = 0 n 1 α s k f ( s k ) h ( g ) = k = 0 n 1 α s k g ( s k )

since for each term where s k s ( f ) we have that f ( s k ) = 0 so that the term contributes nothing to the sum by Definition 6.5.6a (and similarly for g ). Also since f ( ξ 0 ) < g ( ξ 0 ) is has to be that 0 < g ( ξ 0 ) so that ξ 0 s ( g ) and hence ξ 0 S . From this it follows that there is an m < n such that s m = ξ 0 .

We then have

α s m f ( s m ) + k = 0 m 1 α s k f ( s k ) < α s m f ( s m ) + α s m 1 + 1 (by Lemma 6.5.4a and the above) α s m f ( s m ) + α s m (since  { s k }  is an increasing sequence) = α s m [ f ( s m ) + 1 ] (by Definition 6.5.6b) α s m g ( s m ) (since  f ( s m ) < g ( s m ) ) α s m g ( s m ) + k = 0 m 1 α s k g ( s k ) . (by Lemma 6.5.4)

Thus it follows yet again from Lemma 6.5.4a that

h ( f ) = k = 0 n 1 α s k f ( s k ) = k = m + 1 n 1 α s k f ( s k ) + α s m f ( s m ) + k = 0 m 1 α s k f ( s k ) < k = m + 1 n 1 α s k g ( s k ) + α s m g ( s m ) + k = 0 m 1 α s k g ( s k ) = k = 0 n 1 α s k g ( s k ) = h ( g ) ,

which shows that h is indeed increasing.

Lastly we show that h is a surjective function, which then completes the proof that h is an isomorphism so that ( S ( β , α ) , ) is isomorphic to ( α β , < ) as desired. So consider any γ α β so that by definition γ < α β .

We construct a function f : β α recursively by first initializing f ( ξ ) = 0 for all ξ β . We then initialize ρ = γ and perform the following iteratively:

If ρ < α then we set f ( 0 ) = ρ , noting that by definition ρ α and 0 β since β > 0 . We then terminate the iteration, noting that it could be that f ( 0 ) = ρ = 0 .

If ρ α then 1 < α ρ so that by Lemma 6.6.2 there is a greatest ordinal σ such that α σ ρ . Then clearly we have α σ ρ γ < α β so that it follows from Exercise 6.514b that σ < β since α > 1 (the exercise only asserts implication in one direction but the other direction follows immediately similarly to the proof of Lemma 6.5.4a). Then since clearly α σ 0 it follows from Theorem 6.6.3 that there are unique ordinals τ and ρ such that

α σ τ + ρ = ρ

and ρ < α σ . We claim the following about these:

1.
τ < α . To the contrary, assume that τ α so that we have ρ = α σ τ + ρ α σ τ + 0 = α σ τ α σ α = α σ + 1

by Lemma 6.5.4, Exercise 6.5.7 since α σ 0 , and Definition 6.5.9b. However, this contradicts the definition of σ , i.e. that it is greatest ordinal ξ such that ρ α ξ . Hence it must be that τ < α .

2.
0 < τ . Were it the case that τ = 0 then we would have ρ = α σ τ + ρ = α σ 0 + ρ = 0 + ρ = ρ .

However, we then would have both ρ α σ and ρ = ρ < α σ , which is clearly a contradiction. So it must be that τ 0 so τ > 0 .

3.
ρ < ρ . We have ρ < α σ = α σ 1 α σ τ = α σ τ + 0 α σ τ + ρ = ρ

by Exercise 6.5.7 since α σ 0 and Lemma 6.5.4 since 0 ρ , noting that we also just showed that 1 τ since 0 < τ .

We therefore set f ( σ ) = τ , noting that we have shown that σ < β and τ < α (so that σ β and τ α ). We then repeat the above, setting ρ as the new ρ , noting that ρ < ρ γ so that ρ γ is still true.

Now it has to be that this construction terminates after a finite number of iterations since each ρ in the iterations forms a strictly decreasing sequence of ordinals. Hence this sequence has to be finite since otherwise the range of this sequence would be a set of ordinals with no least element. It thus follows that f ( ξ ) 0 for a finite number of ξ β so that s ( f ) is finite and f S ( β , α ) . The fact that h ( f ) = γ follows immediately from the construction of f so that h is surjective since γ was arbitrary. □

In conclusion we note that h maps f S ( β , α ) to its corresponding ordinal γ α β by expanding γ in base- α with digits in the range of f . Running through the above procedures for finite α and β shows that this is the usual base expansion in base- α . However, this is much more general since α or β (or both) might be transfinite. The interesting conclusion here is that any ordinal can be expressed with a finite number of (potentially transfinite) digits with respect to any other ordinal (potentially transfinite) as a base. It would seem that Theorem 6.6.4 (the normal form) is a special case of this in which the base is ω .

User profile picture
2024-07-15 11:42
Comments