Exercise 6.5.11

Find a set A of rational numbers such that ( A , 𝑸 ) is isomorphic to ( α ) where

(a) α = ω + 1 ,

(b) α = ω 2 ,

(c) α = ω 3 ,

(d) α = ω ω ,

(e) α = 𝜀 .

[Hint: { n 1 m m , n 𝑵 { 0 } } is isomorphic to ω 2 , etc.]

Answers

Lemma 1. If β is an initial segment of an ordinal α then β is also an ordinal.

Proof. By Lemma 6.1.2 there is an a α such that β = { x α x /mo> a } . Also, since a α and α is an ordinal, a is an ordinal as well by Lemma 6.2.8. Finally, by the comments after Theorem 6.2.10, β is simply the ordinal a since β = { x α x /mo> a } and each x in that set is an ordinal (by Lemma 6.2.8 since each x α and α is an ordinal). □

Lemma 2. If A is a set of ordinals then ordinal α = sup A if and only if α is the least upper bound of A , i.e. α is an upper bound of A and β is not an upper bound of A for any β /mo> α .

Proof. ( ) First suppose that α = sup A . Then by the remarks following the proof of Theorem 6.2.6 in the text α is an upper bound of A and if β is an upper bound of A then α β . This last statement is simply the contrapositive of the statement that β /mo> α implies that β is not an upper bound of A and hence is logically equivalent.

( ) We show that an ordinal α with the least upper bound property for A is unique, which suffices to show the result since if β has this property then β = sup A since sup A does as well (by what was just shown above) and the ordinal having this property is unique.

So suppose that ordinals α and β both have the least upper bound property for A but that α β . Without loss of generality we can assume then that α /mo> β . But then, since β has the least upper bound property, α cannot be an upper bound of A , which contradicts the fact that α also has the least upper bound property! Hence it has to be that α = β , which shows the uniquness. □

Next we need to build up a little theory.

Definition 1. For an ordinal α we call a set E α an embedding of α if it has the following properties:

1.
E α is a subset of Q .
2.
E α is order isomorphic to α under the usual ordering of the rationals.
3.
For some a and b in 𝑸 , a x /mo> b for every x E α . We denote this by saying a E α /mo> b or that E α is an embedding in [ a , b ) .

We call U α a unit embedding of α if it is an embedding of α in [ 0 , 1 ) .

Theorem 3. If E α is an embedding of α in [ a , b ) then for each x E α there is a Δ x 𝑸 + such that x + Δ x b and if y 𝑸 and x /mo> y /mo> x + Δ x then y is not in E α . That is, x is not a limit point from the right.

Proof. Consider any x E α , noting that x /mo> b .

If x is the greatest element of E α then let Δ x = b x , noting that Δ x /mo> 0 since b /mo> x . We also have

x + Δ x = x + b x = b b .

Then consider any y 𝑸 where x /mo> y /mo> x + Δ x so that clearly y E α since otherwise x would not be the greatest element of E α .

On the other hand if x is not the greatest element of E α then let f be the isomorphism between α and E α and let β = f 1 ( x ) . It follows that β is not the greatest element of α so that β + 1 α as well. Then let Δ x = f ( β + 1 ) x , noting that Δ x /mo> 0 since f ( β + 1 ) /mo> f ( β ) = x since f is an isomorphism. We also have

x + Δ x = x + f ( β + 1 ) x = f ( β + 1 ) /mo> b

since f ( β + 1 ) E α . Now consider any y 𝑸 where f ( β ) = x /mo> y /mo> x + Δ x = f ( β + 1 ) . If it were the case that y E α then we would have that β /mo> f 1 ( y ) /mo> β + 1 since f is an isomorphism, which is impossible since β is an ordinal. So it must be that y E α as desired. □

Corollary 4. If E α is an embedding and x and y are in E α where x /mo> y then x + Δ x y , where Δ x 𝑸 + is that guaranteed by Theorem  3 .

Proof. If it were the case that x + Δ x /mo> y then we have that x /mo> y /mo> x + Δ x , which is in direct contradiction to Theorem  3 since y E α . □

For an embedding E α consider p 𝑸 + and q 𝑸 . We define a set denoted by p E α + q to be the set { 𝑝𝑥 + q x E α } .

Theorem 5. If E α is an embedding in [ a , b ) then F α = p E α + q is also an embedding of α for any p 𝑸 + and q 𝑸 . Moreover 𝑝𝑎 + q F α /mo> 𝑝𝑏 + q .

Proof. So first consider any y F α so that y = 𝑝𝑥 + q for some x E α . Since E α is an embedding x 𝑸 so that clearly y = 𝑝𝑥 + q 𝑸 as well since p , q 𝑸 . Hence since y was arbitrary we have that F α 𝑸 so that (1) is satisfied.

Now consider the mapping f : E α F α defined by f ( x ) = 𝑝𝑥 + q for x E α . Clearly F α = { f ( x ) x E α } so that f is onto. Consider then x , y E α where x /mo> y so that we have

x /mo> y 𝑝𝑥 /mo> 𝑝𝑦 (since  p /mo> 0 ) 𝑝𝑥 + q /mo> 𝑝𝑦 + q f ( x ) /mo> f ( y ) .

Hence f is an isomorphism. Thus F α is isomorphic to E α so that clearly it is also isomorphic to α since E α is, thereby showing (2).

Lastly for any y F α we have y = 𝑝𝑥 + q for some x E α . We then have

a x /mo> b 𝑝𝑎 𝑝𝑥 /mo> 𝑝𝑏 (since  p /mo> 0 ) 𝑝𝑎 + q 𝑝𝑥 + q /mo> 𝑝𝑏 + q 𝑝𝑎 + q y /mo> 𝑝𝑏 + q .

This shows both (3) and the last statement. □

For an embedding E α in [ a , b ) and a unit embedding U β for ordinals α and β we define the product

E α U β = x E α ( Δ x U β + x ) ,

where Δ x 𝑸 + is that guaranteed to exist by Theorem 3.

Theorem 6. For an embedding E α in [ a , b ) and a unit embedding U β the product E α U β is an embedding of β α in [ a , b ) .

Proof. First consider any y E α U β so that there is an x E α such that y Δ x U β + x . Since y Δ x U β + x is an embedding by Theorem 5 it follows that y 𝑸 , which shows (1) since y was arbitrary.

Now since E α is an embedding of α there is an isomorphism f : α E α . Similarly there is an isomorphism g : β U β since U β is an embedding of β . Consider α × β with lexicographic ordering . Then define a mapping h : α × β E α U β by

h ( δ , 𝜀 ) = Δ f ( δ ) g ( 𝜀 ) + f ( δ )

for ( δ , 𝜀 ) α × β , noting that Δ f ( δ ) is that guaranteed by Theorem 3 since f ( δ ) E α .

First we claim that h is surjective. So consider any z E α U β so that there is an x E α such that z Δ x U β + x . By definition then there is a y U β such that z = Δ x y + x . Now let δ = f 1 ( x ) and 𝜀 = g 1 ( y ) so that x = f ( δ ) and y = g ( 𝜀 ) , which can be done since f and g are bijections. We then have that

h ( δ , 𝜀 ) = Δ f ( δ ) g ( 𝜀 ) + f ( δ ) = Δ x y + x = z ,

which shows that h is surjective since z was arbitrary.

We also claim that h is an isomorphism and therefore also injective. So consider any ( δ 1 , 𝜀 1 ) and ( δ 2 , 𝜀 2 ) in α × β where ( δ 1 , 𝜀 1 ) ( δ 2 , 𝜀 2 ) . By the definition of lexicographic ordering we have the following:

Case: δ 1 /mo> δ 2 . Then since f is an isomorphism f ( δ 1 ) /mo> f ( δ 2 ) , and also by Corollary 4 it follows that f ( δ 1 ) /mo> f ( δ 1 ) + Δ f ( δ 1 ) f ( δ 2 ) . We also have that

Δ f ( δ 1 ) g ( 𝜀 1 ) + f ( δ 1 ) /mo> Δ f ( δ 1 ) + f ( δ 1 )

by Theorem 5 since g ( 𝜀 1 ) U β and U β is a unit embedding. Also since g ( 𝜀 2 ) 0 (since g ( 𝜀 2 ) U β ) and Δ f ( δ 2 ) /mo> 0 (by Theorem  3) that

f ( δ 2 ) Δ f ( δ 2 ) g ( 𝜀 2 ) + f ( δ 2 ) .

Combining all this results in

h ( δ 1 , 𝜀 1 ) = Δ f ( δ 1 ) g ( 𝜀 1 ) + f ( δ 1 ) /mo> Δ f ( δ 1 ) + f ( δ 1 ) f ( δ 2 ) = h ( δ 2 , 𝜀 2 ) .

Case: δ 1 = δ 2 and 𝜀 1 /mo> 𝜀 2 . Then obviously f ( δ 1 ) = f ( δ 2 ) so that Δ f ( δ 1 ) = Δ f ( δ 2 ) but also g ( 𝜀 1 ) /mo> g ( 𝜀 2 ) since g is an isomorphism. We then have

g ( 𝜀 1 ) /mo> g ( 𝜀 2 ) Δ f ( δ 1 ) g ( 𝜀 1 ) /mo> Δ f ( δ 1 ) g ( 𝜀 2 ) (since  Δ f ( δ 1 ) /mo> 0 ) Δ f ( δ 1 ) g ( 𝜀 1 ) + f ( δ 1 ) /mo> Δ f ( δ 1 ) g ( 𝜀 2 ) + f ( δ 1 ) Δ f ( δ 1 ) g ( 𝜀 1 ) + f ( δ 1 ) /mo> Δ f ( δ 2 ) g ( 𝜀 2 ) + f ( δ 2 ) (since  Δ f ( δ 1 ) = Δ f ( δ 2 )  and  f ( δ 1 ) = f ( δ 2 ) ) h ( δ 1 , 𝜀 1 ) /mo> h ( δ 2 , 𝜀 2 ) .

Thus in all cases h ( δ 1 , 𝜀 1 ) /mo> h ( δ 2 , 𝜀 2 ) , which shows that h is an isomorphism since ( δ 1 , 𝜀 1 ) and ( δ 2 , 𝜀 2 ) were arbitrary. Hence E α U β is isomorphic to the lexicographic ordering of α × β and therefore also to β α by Theorem 6.5.8. This shows part (2) of the embedding definition.

Lastly consider any z E α U β so that there is an x E α such that z Δ x U β + x . Then since U β /mo> 1 we have that z /mo> Δ x + x b by Theorems 5 and 3. Also since 0 U β it follows from Theorem 5 that a x z since a E α . Since z was arbitrary this shows that a E α U β /mo> b , which shows (3). This completes the proof. □

Theorem 7. Suppose that α is a limit ordinal and that { α n } is a sequence ( n ω ) of nonzero ordinals in α . Also suppose that E ω is an embedding of ω in [ a , b ) and U α n is a unit embedding of α n for each n ω . Suppose further that α = sup n ω α n and that α n + α n + 1 = α n + 1 for every n ω , noting that clearly n + 1 ω as well. Lastly, suppose that f is the isomorphism from ω to E ω . Let

A n = Δ f ( n ) U α n + f ( n )

for n ω . Then the set

E α = n ω A n

is an embedding of α in [ a , b ) .

Proof. First consider any n and m in ω where n m . We can assume without loss of generality n /mo> m . Then f ( n ) /mo> f ( m ) since f is an isomorphism and, moreover, it follows from Corollary 4 that Δ f ( n ) + f ( n ) f ( m ) . Hence by Theorem 5 we have that

A n = Δ f ( n ) U α n + f ( n ) /mo> Δ f ( n ) + f ( n ) f ( m ) Δ f ( m ) U α m + f ( m ) = A m

since U α n and U α m are unit embeddings. Hence all the A n are disjoint and moreover A 0 /mo> A 1 /mo> A 2 /mo> . Also clearly by Theorem  5 each A n is isomorphic to α n since U α n is.

Now for p , q 𝑸 let [ p , q ) = { x 𝑸 q x /mo> q } . We then claim that E α [ a , f ( n + 1 ) ) for any n ω is isomorphic to α n , which we shall show by induction on n . For n = 0 we clearly have that a A 0 Δ f ( 0 ) + f ( 0 ) f ( 1 ) A m for any m 1 . From this it follows that E α [ a , f ( n + 1 ) ) = E α [ a , f ( 1 ) ) = A 0 , which is isomorphic to α 0 = α n by what was shown above.

Now suppose that E α [ a , f ( n + 1 ) ) is isomorphic to α n . We clearly have E α [ a , f ( n + 2 ) ) = [ E α [ a , f ( n + 1 ) ) ] [ E α [ f ( n + 1 ) , f ( n + 2 ) ) ] and that E α [ f ( n + 1 ) , f ( n + 2 ) ) = A n + 1 , which is isomorphic to α n + 1 by what was shown above. Then E α [ a , f ( n + 2 ) ) is the sum of E α 𝑐𝑙𝑜𝑝𝑎 , f ( n + 1 ) and E α [ f ( n + 1 ) , f ( n + 2 ) ) = A n + 1 so that by Theorem 6.5.3, the induction hypothesis, and the given property of { α n } it is isomorphic to α n + α n + 1 = α n + 1 . This completes the inductive proof.

We also show that E α [ a , f ( n + 1 ) ) is an initial segment of E α for any n ω . So consider any such n , any x E α [ a , f ( n + 1 ) ) , and any y E α where y /mo> x . Since a E α clearly a y . We also have y /mo> x /mo> f ( n + 1 ) (since x [ a , f ( n + 1 ) ) ). Hence y [ a , f ( n + 1 ) ) so that also y E α [ a , f ( n + 1 ) ) , which shows that E α [ a , f ( n + 1 ) ) is an initial segment of E α by definition.

Now we claim that E α is a well-ordered set. So consider any nonempty subset B of E α . Then there is some x B and since x E α there is an n ω such that x A n . Then clearly x B [ a , f ( n + 1 ) ] so that B [ a , f ( n + 1 ) ] is a nonempty subset of E α [ a , f ( n + 1 ) ) . It was shown above that E α [ a , f ( n + 1 ) ) is isomorphic to α n so that it is a well-ordered set. Hence B [ a , f ( n + 1 ) ] has a least element y . We claim that this is the least element of B , so consider any z B . If z /mo> f ( n + 1 ) then clearly z B [ a , f ( n + 1 ) ) so that obviously y z since y is the least element of B [ a , f ( n + 1 ) ) . On the other hand if z f ( n + 1 ) then y /mo> f ( n + 1 ) z (since y [ a , f ( n + 1 ) ) ) so that again y z . Since z was arbitrary this shows that y is in fact the least element of B . Since B was an arbitrary subset of E α this shows that E α is well-ordered.

Since E α is a well-ordered set it is isomorphic to some ordinal γ by Theorem 6.3.1. We then claim that γ = α . Letting C = { α n n ω } , we show this by showing that γ is the least upper bound of C , which shows that γ = α by the least upper bound property (Lemma 2) since α = sup C by definition. So first consider any α n C . It was shown above that α n is isomorphic to E α [ a , f ( n + 1 ) ) , and this was shown to be an initial segment of E α , which itself is isomorphic to γ . Thus it follows that that α n /mo> γ so that α n γ is true. Since α n was arbitrary this shows that γ is an upper bound of C .

Now consider any ordinal δ /mo> γ so that δ γ . Let g be the isomorphism from γ to E α since it has been shown that they are isomorphic. Then since g ( δ ) E α so that there is an n ω such that g ( δ ) A n . Then also g ( δ ) E α [ a , f ( n + 1 ) ) . Since E α [ a , f ( n + 1 ) ) is an initial segment of E α (just shown above) it follows that g 1 [ E α [ a , f ( n + 1 ) ) ] is an initial segment of γ . Since γ is an ordinal it follows from Lemma  1 that g 1 [ E α [ a , f ( n + 1 ) ) ] is an ordinal. Since g 1 E α [ a , f ( n + 1 ) ) is an isomorphism (since g is) and E α [ a , f ( n + 1 ) ) is isomorphic to α n (shown above), it follows that g 1 [ E α [ a , f ( n + 1 ) ) ] is in fact α n ! Then, since g ( δ ) E α [ a , f ( n + 1 ) ) we have that δ g 1 [ E α [ a , f ( n + 1 ) ) ] = α n so that δ /mo> α n . Hence δ is not an upper bound of C . Since δ was arbitrary this shows that γ is in fact the least upper bound of C so that γ = α .

Parts (1) and (3) of the definition of an embedding are trivial to show by the same arguments as those used in the proof of Theorem 6. Hence E α is an embedding of α in [ a , b ) . □

Main Problem.

(a) Define f : ω 𝑸 by

f ( n ) = 1 1 n + 1 = n n + 1

for n ω . Then let U ω = { f ( n ) n ω } . We claim that U ω is a unit embedding of ω .

Proof. Clearly U ω 𝑸 so that (1) is satisfied. To show (2) let g : ω U α be defined by g ( n ) = f ( n ) = 1 1 ( n + 1 ) , which is clearly onto based on the definition of U α . We also claim that g is an isomorphism. To this end consider any n , m ω where n /mo> m . We then have

n /mo> m n + 1 /mo> m + 1 n + 1 m + 1 /mo> 1 (since  m + 1 1 /mo> 0  since  m 0 ) 1 m + 1 /mo> 1 n + 1 (since  n + 1 1 /mo> 0  since  n 0 ) 1 m + 1 /mo> 1 n + 1 1 1 m + 1 /mo> 1 1 n + 1 g ( m ) /mo> g ( n ) g ( n ) /mo> g ( m ) .

Hence g is an isomorphism so that U ω is indeed isomorphic to ω , which shows (2).

Lastly consider any x U ω so that x = f ( n ) = 1 1 ( n + 1 ) for some n ω . Then we have

n 0 /mo> 1 n + 1 1 /mo> 0 1 1 n + 1 /mo> 0 (since  n + 1 1 /mo> 0 ) 1 1 n + 1 /mo> 0 0 1 1 n + 1 /mo> 1 0 x /mo> 1 .

Since x was arbitrary this shows that 0 U ω /mo> 1 so that (3) holds and U ω is a unit embedding.

Now let E 1 = { 1 } . Clearly this is an embedding of the ordinal 1. Moreover we have 0 U ω /mo> 1 E 1 /mo> 2 so that U ω and E 1 are disjoint. Then clearly U ω E 1 is the sum of U ω and E 1 so that it is isomorphic to ω + 1 by Theorem 6.5.3 and thus an embedding of ω + 1 since it is trivial to show that 0 U ω E 1 /mo> 2 . □

(b) Now consider the same U ω from part (a) and let E ω = 1 U ω + 1 so that by Theorem 5 this is another embedding of ω and 1 E ω /mo> 2 . Hence we have that 0 U ω /mo> 1 E ω /mo> 2 so that U ω and E ω are disjoint. Also clearly E ω 2 = U ω E ω is the sum of U ω and E ω so that it is isomorphic to ω + ω = ω 2 by Theorem 6.5.3. Hence since also 0 E ω 2 /mo> 2 (it is trivial to show) it is an embedding of ω 2 as desired.

(c) Considering the same E ω 2 from part (b) let F ω = 1 U ω + 2 so that it is yet another an embedding of ω and 2 F ω /mo> 3 by Theorem  5. Then by the same arguments as in part (b) it follows that E ω 2 F ω is an embedding of ω 2 + ω = ω 2 + ω 1 = ω ( 2 + 1 ) = ω 3 as desired.

(d) We first aim to construct a unit embedding of ω n for any n ω . We do this recursively. For n = 0 clearly U ω 0 = { 0 } is a unit embedding of ω 0 = 1 . We have also already constructed U ω as a unit embedding of ω in part (a). Now suppose we have constructed U ω n , a unit embedding of ω n . Then we let U ω n + 1 = U ω U ω n so that U ω n + 1 is a unit embedding of ω n ω = ω n + 1 by Theorem 6.

Clearly then { ω n } is a sequence of ordinals in ω ω and by definition ω ω = sup n ω ω n . We also have in hand the embedding U ω and U ω n for each n ω as just constructed recursively. Now consider any n ω so that we have we have

ω n + ω n + 1 = ω n 1 + ω n ω = ω n ( 1 + ω ) = ω n ω = ω n + 1 .

We can therefore apply Theorem 7 to construct a unit embedding U ω ω of ω ω .

(e) Here we use the operation of tetration, which we define recursively for all ordinals. So for all ordinals β we define:

1.
0 β = 1
2.
α + 1 β = β α β for all α
3.
α β = sup { γ β γ /mo> α } for all limit α 0 .

Thus we have 1 ω = ω , 2 ω = ω ω , 3 ω = ω ω ω , etc. We then have that 𝜀 = ω ω = sup n /mo> ω n ω by definition.

Despite working and thinking about this problem for weeks I have been unable to come up with a way to embed 𝜀 . To be sure we can easily construct embeddings of ordinals larger than ω ω , for example we can embed ω ω ω ω = ω ω + ω = ω ω 2 by simply applying Theorem 6 to our just-constructed embedding of ω ω . However, I could think of no way to get to 𝜀 easily using this approach. Ideally what we would have is a way to construct an embedding of ω α for any ordinal α for which we already have an embedding. This would easily lead to an embedding of 𝜀 using Theorem 7 after constructing the embeddings for n ω recursively. However, I was unable to think of how to construct this and eventually had to admit defeat and move on.

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