Chapter 9 Cantor Normal Forms

Chapter 9 Cantor Normal Forms

CHAPTER 9 CANTOR NORMAL FORMS 9.1 In this chapter we prove the existence of Cantor Normal Formsg1 for a large class of co-ordinals, but first of all...

272KB Sizes 2 Downloads 115 Views

CHAPTER 9

CANTOR NORMAL FORMS

9.1 In this chapter we prove the existence of Cantor Normal Formsg1 for a large class of co-ordinals, but first of all we prove a general decomposition theorem for all co-ordinals less than some principal number for addition. We restrict our attention from now until the end of Part One of this monograph to co-ordinals, returning to the consideration of quords in Part Two. Since we are dealing with co-ordinals we shall repeatedly use the fact (theorem 4.2.2) that A < B e ( 3 C )( A + C = B & C

=+ 0).

9.1.1 LEMMA. A co-ordinal P is a principal number for addition if, and only if, A
9.1.2 THEOREM. If P E S (+), then P. WE&( +) and there is no principal number Q such that P
PROOF. Suppose P E X (+) and A < P -W ,then by theorem 6.2.5, there exist n, D such that

A = P * n + D , where D < P.

Ch. 91

111

CANTOR NORMAL FORMS

Hence A + P . W = A +(P + P - W ) =(A +P)+P-W =(Pen D +P)+P-W = P . ( n + 1) + P a w =P.W.

since 1 + W = W ,

+

sincePsX(+)

and D < P ,

Hence P.WE*( +). Clearly PCP-W since P 9 0. Now suppose Q s X ( +) and P < Q , then by lemma 9.1.1, P*W
hence Q


9.1.3 THEOREM. g2 If O c A < P E X ( +), then there exist principal numbers for addition Ply ..., Pnsuch that P > P,,2 . . * 2Pl

and A = P,,

Pl.

A = Q,

Q1

+ a * * +

Further, if + * * a +

is any other representation of A as a sum of principal numbers Q, (i=l,..., m) such that

Q,

>***

2

Q1,

then m = n & P i = Qi for i

=

1, ..., n.

Conversely, if A is expressible in the form (2) with P,, 2

2 PI where the

Pi(i= 1, ..., n) are principal numbers for addition, then there is a principal number P e.g. P,,. W,such that P > A and P > Pi for i

=

1, ..., n.

PROOF(by transfinite induction on the partial well-ordering I). We assume O < A < P E H ( + ) and take as induction hypothesis: If O c B c A

112

CONSTRUCTIVE ORDER TYPES

[Ch. 9

then B is uniquely expressible in the form (2) where the P, are principal numbers satisfying (1). If A is a principal number for addition then the theorem follows at once from theorem 9.1.2. Now suppose A is not a principal number, then there exist B, C such that B+C=A,C+O,A

and B + C < P .

(3)

By corollary 8.1.6, C
Let P, be the least C satisfying (3). PI exists since { C :c < A }

is well-ordered by I .We show that P, is a principal number for addition. Suppose Pi = D + E , then by corollary 8.1.6, E c P and hence by theorem 4.2.8, PI and E are comparable. But

(El5 lP1l

I

hence E IPi

and by the minimality of P,, E=P,.

Therefore PIis a principal number by theorem 8.1.2. Now let B, be the least B such that B + P, = A . If B, = 0, then A is a principal number which contradicts our assumption. Hence B, + O

and by the induction hypothesis B, has a unique decomposition where

Ch. 91

113

CANTOR NORMAL FORMS

and all the Pi(i = 2, ..., n) are principal numbers. Therefore A = Pn +**.+Pl

and, since p, < p , all the Pi (i = 1, ..., n) are comparable by theorem 4.2.8. Suppose pz < PlB

then, since Pl is a principal number for addition,

Pz +PI =PI; hence A=(Pn+-.+Pz)+P1 =(P.+--+P,)+P,.

Now

*

Pz 0 3 Pn+

9.-

+ P3 < P, + - + P2 = B , , **

whereas Bl was chosen as the least B such that B + P , = A . This contradiction shows that Pz 2 PI and we conclude that A has a decomposition of the form (2). As regards uniqueness, let A

= P,, +..,+PI

and A = Q,

+-.+ Q ,

be two decompositions of A as a sum of non-increasing principal numbers. By theorem 4.2.8, P,,and Q, are comparable. Suppose Pn>Qm,

then, since P,,is a principal number for addition,

Q, Therefore

+ P,,= P,.

+ P,, Pi = Q , + Q , + P,, - ... = Q,.(m + 1) + A .

A = Q,

+ a * * +

+ * a * +

Since A+O, we therefore have Q,.(m

+ 1) < A .

114

[Ch. 9

CONSTRUCTIVE ORDER TYPES

On the other hand, if i Im, then or Qm.2,

Qi+Qm=Qm

according as Qi

< Qm

or

= Qm.

Qi

Consequently, we have Q,.(m

+ 1) + A = A < A + Q;m

5 Q;2m

and by corollary 3.2.9, it follows that A
(5)

which contradicts (4) above. We conclude Q, 4: P,, and similarly P,, 4: Q,,

but since P,, and Q, are comparable it follows that

P,,= Q, and P,-,

+...+ P, = Q,-, +...+ Q1

(by corollary 3.2.9). Proceeding thus we obtain after s( = minimum of m, n ) steps P,,-r = Qm-r r = 0, ..., s and either P,+ - . , + P I= 0 or Q, Q, = 0 , + . e m +

where t = maximum of m - s , n-s. By theorem 2.2.4.(ii) it follows that t =0 and hence that m =n and

Pi= Qi for i

= 1, ..., n ( = m )

For the converse: if A=P,,+**.+P,,

then by the argument which produced ( 5 ) we have A I P;n

and consequently A < P,,. W .

By theorem 9.1.2, P,,. Wis a principal number for addition. This completes the proof.

Ch. 91

115

CANTOR NORMAL FORMS

9.2.

LEMMA. C
9.2.1

The next lemma is a strengthening of theorem 6.2.5. 9.2.2

LEMMA.If A 90,then

c < A B o ( 3 D ) ( D < B & A D 5 c < A ( D + 1)). PROOF.The implication from right to left is clear from corollary 5.2.3, theorem 8.4.1 and the transitivity of < for co-ordinals. Suppose now that C
then there exists FPO such that C+F=AB.

Given AEA and BEB, then by the separation lemma 2.3.1, there exists CEC and F E F such that [C)(Fand]

C $ F=A.B.

Let

f = min(F), d = l(f), D = d) B and D

= COT(D).

We shall show that for this D, ADIC
Now AD E C E AB,

since j ( a , x ) ~ e A D = . xsB d =>j(a7 x )

SABf

=.j(a, X)EC'C.

Further, since AD and C are both initial segments of AB, we must have ADsC. By corollary 5.2.3,

D
116

[Ch. 9

CONSTRUCTIVE ORDER TYPES

and hence by theorem 8.4.1.(ii) it follows that A(D

+ 1) IA B .

From theorem 4.2.8 we now have or C I A ( D + l ) ,

A(D+1)1C

but

c c A * ( DT ( ( 4d ) ) ) and so we have

c IA ( D + 1).

Finally, since

f E C j ( A , d ) - C ' C , C + A ( D + 1) and the proof is complete. 9.2.3

LEMMA.1 1 C < A ~ o ( 3 0 ) ( D < B & A D s C < A D f l ) .

PROOF. The implication from right to left follows at once from corollary 5.2.3, theorem 8.5.5.(ii) and the transitivity of <. Now suppose lIC
C +F

=AB.

Further, there exist A, B such that

A = [AIEA and B = [B]EB and by the separation lemma 2.3.1, there exist C E Cand FEFsuch that

c~

F=A~.

Let f = min(F),

then f

=

( b o ... b:) a,

where bo>

... > b,,,b,eB

... a,

(i=O,...,n), n+-1

(since C21) and93

O+U,EA( i = O , . . . , H).

Also let

D = bo)B and D = COT (D) ,

Ch. 91

CANTOR NORMAL FORMS

117

then we claim ADs C. Now XEC'A" implies

where do
f,

which implies XEC'C. Further, if y€C'ADand < y , x)eAD, then by the same argument Y < A exp B f It follows at once (from corollary 4.2.5) that AD Ic.

Now let E = D T ((bo, bo)} ,

then XEC'CJX=O

or x = e

where bb>-..>b;, biEB and 0 8 a ; e A . Further, by the definition of bo, (bb, bo)

E

B

and hence x€C'AEand

C5AEdD+'. But C 8 AEsince

f E C'AE - C'C and it follows that

AD 5 C < A D + ' . LEMMA. A T W & B< C*AB+ AC= A'. PROOF. Since B < C , (30)( D + O & B + D = C ) .

9.2.4

Hence + A =~ A +~A ~ . A ~ = AB(l + A D ) = AB(l + A E)

+

by theorem 7.3.4, by theorem 6.2.1, by lemma 8.5.6.(i) , where A + E = A D ,

118

[Ch. 9

CONSTRUCTIVE ORDER TYPES

+ W + F + E)

= A'(1

= AB(W

+ F + E)

=A ~ A D =A

9.2.5

+

since W < A* W F = A for some F , by theorem 5.2.6.(i),

~ .

LEMMA. If C < AB, then there exist unique Q, R such that C = AQ

+R ,

where 0 I Q < B & O I R < A .

PROOF. By lemma 9.2.2, C < AB*(3Q)(O I Q < B & AQ I C < A ( Q

Hence AQ I C < AQ

+ 1)).

+A .

Since A , C, Q are co-ordinals, R=C-AQ

+R as required. Now suppose C = AQ + R = AQ, + R , ,

is well-defined, 0 5 R < A and C = A Q

where 0 1 Q,
and O I R < A ,

by theorem 4.2.8 Q and Q, are comparable (and so too are R and R,). By corollary 5.2.3, if Q + Q, then either Q 1 + l I Q or Q + l S Q , .

Suppose the former holds, then AQl I C < A (Q1

+ 1) I AQ < C ,

which is a contradiction. Similarly we cannot have Q

+ 1s Q,, hence

Q = Qi and by corollary 3.2.9, R=R,. 9.2.6

LEMMA.If 1I C< A* then there exist unique D , Q, R such that

C = A D Q + R , where D < B ,

O
and O j R < A D .

Ch.91

119

CANTOR NORMAL FORMS

PROOF. By lemma 9.2.3, 1 IC < AB* ( 3 0 ) ( A DIC < AD+' & D < B ) , hence C < AD+' = A D . A .

Now, if A < C, D + 0. By lemma 9.2.5, there therefore exist Q, R such that C = A D Q + R , where O < Q < A

and O I R < A .

Suppose that we also had C = AEQ, + R , ,

where E < B ,

O
and O I R , < A .

By theorem 4.2.8, D and E are comparable and hence by corollary 5.2.3, if D =kE, then D + l I E or E + l l D .

Suppose the former holds, then by theorem 8.5.5.(i), and using the fact that A E is a co-ordinal, we have A ~ cS < A ~ + I '

C,

which is a contradiction. Likewise we cannot have E + l s D and we conclude D=E.

Now suppose A 2 C, then either trivially A = C or C = A o Q + R , where O < Q < A

and R = O

and Q, R are unique by the preceding lemma. Thus for all A satisfying the hypothesis we obtain D, Q, R uniquely as required. 9.2.7

THEOREM. If 1 IC < AB then C is uniquely expressible in the form C = A B 1 * D+ 1 .-.+ ABr.D,,

(1)

where B > B , > - . . > B , and l l D , < A f o r l l i l r . PROOF. By lemma 9.2.6, 1 Ic < A~ 3 c

=

D,

+ c, ,

where O I D , < A , O I C , < A B ' and B , < B and B,, D,, C, are unique. Since Cl < AB' < A B and AB' I C < A B

120

CONSTRUCTWE ORDER TYPES

[Ch. 9

it follows that C, < C < A B .

Since {E:E< A B } is well-ordered by Iwe obtain, by a finite number of repetitions of the above process, the decomposition (1). Suppose that C is also expressible as C = AF'.H1 +-.*+AFs*Hs, where B> F, > >F, and 1IHi
+

C = AB'*D1 C1 = AB1*Hl+ El ,

where 0
or H,ID,+J

D,+IIH, Suppose the former holds, then

C=A~'.D1+C1
which is a contradiction. Similarly, H, $ D, + 1 and we conclude

D, = H , . By corollary 3.2.9 it follows that C, = El

and hence by induction on the maximum of r, s that Bi = Fi and D i = H i for 1 Ii Ir = s .

This completes the proof. As an immediate consequence we obtain our main theorem of this section. 9.2.8 THEOREM (CANTORNORMAL

FORM).

If O
Ch. 91

CANTOR NORMAL FORMS

121

uniquely expressible in the form

c=~ where A > A , >

~ * +...+ - n , WAp*n,,

(2)

> A , and the ni are finite, non-zero co-ordinals.

COROLLARY. A necessary and sufficient condition that a co-ordinal C+O should have a Cantor Normal Form (2) is that there should exist a co-ordinal A such that C < WA. PROOF. The condition is obviously sufficient. Now suppose C has the given Cantor Normal Form and let A be greater than A , (say, take A =A, + 1); then by n, ...+n, applications of lemma 9.2.4, we obtain C + WA=W Awhence C < WA.

9.2.9

+