CODING AND RESHAPING WHEN THERE ARE NO SHARPS

Saharon Shelah1,2,5 and Lee J. Stanley3,4,5

Abstract. Assuming 0 does not exist, is an uncountable cardinal and for all cardinals with < + , 2 = + , we present a mini-coding between and + . This allows us to prove that any subset of + can be coded into a subset, W of + which, further, reshapes the interval [, + ), i.e., for all < < + , = (card )L[W ] . We sketch two applications of this result, assuming 0 does not exist. First, we point out that this shows that any set can be coded by a real, via a set forcing. The second application involves a notion of abstract condensation, due to Woodin. Our methods can be used to show that for any cardinal , condensation for holds in a generic extension by a set forcing.

0. INTRODUCTION. Theorem. Assume that V |= ZF C + 0 does not exist, and, in V, 2 , Z + and for cardinals with < + , 2 = + . THEN there is a conality preserving forcing S() = S(, Z) of cardinality +(+1) such that if G is V -generic for S(), there is W + such that V [G] = V [W ], Z L[W, Z ], for all cardinals with < + , and for all limit ordinals with < < + , = (card )L[W ] . Our forcing S() can be thought of as a kind of Easton product between and + of partial orderings which simultaneously perform the tasks

revision:1996-03-11

modified:1996-03-11

1. Research partially supported by NSF, the Basic Research Fund of the Israel Academy of Science, and MSRI. 2. Paper number 294. 3. Research partially supported by NSF grant DMS 8806536 and MSRI. 4. Preparation of the nal version of this paper partially supported by a grant from the Reidler Foundation. 5. We are grateful to the administration and sta of MSRI for their hospitality during portions of 1989-90. 1

294

SAHARON SHELAH1,2,5 AND LEE J. STANLEY3,4,5

of coding (1.2 of [1]) and reshaping (1.3 of [1]). Our new idea is to introduce an additional coding area used for marking certain ordinals. This marking technique is the crucial addition to the arguments of 1 of [1]. We appeal to the Covering Lemma twice: in (2.1), and again in the proof of the Proposition in (2.3). The referee has informed us that the hypothesis that 0 does not cannot be eliminated. Jensen rst used this hypothesis in [1] to facilitate certain arguments, and then realized that his uses were eliminable. It is not the purpose of this paper to discuss the nature of Jensens appeals to the Covering Lemma; the interested reader may consult pp. 62, 96 and the Introduction to Chapter 8 of [1] for insight into Jensens uses of the Covering Lemma, and how he was able to eliminate them. In [2], S. Friedman presents a rather dierent, more streamlined approach to avoiding such uses of Covering. It should be clear from the preceding that Jensens appeals to the Covering Lemma are of a rather dierent character than ours. To better understand the role of this marking technique, let us briey recall some material from [1]. Let us rst consider the possibility of coding R + into a subset of , when is regular. In order to use almost disjoint set coding, we seem to need extra properties of the ground model, or of the set R, since, in order to carry out the decoding recursion across [, + ) we
modified:1996-03-11

need, e.g., an almost disjoint sequence b = (b : [, + )) of conal subsets of satisfying: () : for all [, + ), (b : ) L[R ], and is canonically denable there. Such a b is called decodable, and it is easy to obtain a decodable b if R satises: () : for all [, + ), (card )L[R] = .

revision:1996-03-11

If () holds, we say that R promptly collapses fake cardinals. Of course, typically () fails, and the reshaping conditions of 1.3 of [1] are introduced to obtain () in a generic extension. Our and R, from the previous paragraph are called and B in 1.3 of [1]. Unfortunately, the distributivity argument for the reshaping partial ordering given there seems to really require not merely that H + = L + [B], but that H ++ = L ++ [B], where B + . This will be the case if B is the result of coding as far as + , but that is another story, which leads to Jensens original approach to the

294

CODING AND RESHAPING WHEN THERE ARE NO SHARPS

modified:1996-03-11

Coding Theorem. Our appeals to the Covering Lemma focus on this point: essentially, to prove a distributivity property of the reshaping conditions. As already indicated, in Jensens treatment, the appeals to the Covering Lemma were designed to overcome dierent sorts of obstacles and proved to be eliminable. Our approach to guaranteeing that the unions of certain increasing chains of reshaping conditions collapse the suprema of their domains is to have marked a conal sequence of small order type. Because of the need to meet certain dense sets in the course of the construction, it is too much to expect that the ordinals we intentionally marked are the only marked ordinals. However, what we will be able to guarantee is that they are the only members of a certain club subset which have been marked. The club will exist in a small enough inner model, thanks to the Covering Lemma. This argument is given in (2.3). We are grateful to the referee for suggesting the use of fast clubs in the argument of (2.3). This allowed us to streamline a more complicated argument (which also suered from some [probably reparable] inaccuracies) in an earlier version of this paper. We use 1 to mark ordinals. To guarantee that this does not collide with requirements imposed by the coding part of the conditions, we set aside the limit ordinals as the only potentially marked ordinals and do not use them for coding. SUMMARY AND ORGANIZATION. We now give a brief overview of the contents of this paper. In 1, we build to the denition, in (1.5), of the S(), along with auxiliary forcings, Sk (). In 2, we prove that the S() are as required. The heart of the matter is (2.3), where we prove the distributivity properties of the Sk (). Preliminary observations are given in (2.1) and (2.2). The former shows that only increasing sequences of certain lengths are problematical. The latter is a rather routine observation about how the coding works. In the argument of (2.3), we use this in the context of forcing over N , a transitive set model of enough ZF C, introduced in the proof of (2.3), below. In (2.4) we put together the material of (2.1) - (2.3) to prove the Theorem. In (2.5) we make a few remarks and briey sketch the applications mentioned in the abstract. The partial ordering S(T, ), introduced in (1.2), below, is the analogue of the reshaping partial ordering of (1.3) of [1]. It adds a subset of + ,

294

revision:1996-03-11

SAHARON SHELAH1,2,5 AND LEE J. STANLEY3,4,5

which, together with T , promptly collapses fake cardinals in (, + ). The partial ordering P, T, g , introduced in (1.4), is a version of the coding partial ordering of (1.2) of [1], relative to g. We require that T + , g S(T, + ). If p P, T, g , then p will have the form ( (p), r(p)); (p) is the function part of p and r(p) is the promise part of p. We require that (p) starts to code not only T , but also g and that (p) S(T , ). If g were not merely a condition but generic for S(T, + ), then P, T, g would just be the usual forcing for the almost-disjoint set coding of the join of T and g, with the extra requirement above, that for conditions, p, (p), together with T , collapses sup dom (p). Finally, the S(), introduced in (1.5), is the forcing which accomplishes the task of coding and reshaping, between and + . It is dened relative to the choice of a xed Z + such that H+n = L+n [Z +n ], for all n . The elements of S(), are -sequences, (p(n) : n < ), where for all n < , p(n) = ( (p(n)), r(p(n))), (p(n)) S(Z n , n ) and p(n) Pn , Zn+1 , (p(n+1)) . Thus, letting G be the canonical name for the generic of S(), letting G(n) be the canonical name for { (p(n)) : p G}, and letting W (n) be the canonical name for G(n), S() is a sort of Easton product of the Pn ,
modified:1996-03-11

Zn+1 , W (n+1) .

NOTATION AND TERMINOLOGY. Our notation and terminology is intended to be standard, or have a clear meaning, e.g., o.t. for order type, card for cardinality. A catalogue of possible exceptions follows. When forcing, p q means q gives more information. Closed unbounded sets are clubs . The set of limit points of a set X of ordinals is denoted by X . AB is the symmetric dierence of A and B, and A\B is the relative complement of B in A. For ordinals, , [, ) is the half-open interval { : < }. The notation for the three other intervals are clear. It should be clear from context whether the open interval or the ordered pair is meant. OR is the class of all ordinals. For innite cardinals, , H is the set of all sets hereditarily of cardinality < , i.e. those sets x such that if t is the transitive closure of x, then card t < . For ordinals , , we write >> to mean that is MUCH greater than ; the precise sense of how much greater we must take it to be is supposed to be clear from context. For models, M, SkM denotes the Skolem operation in M, where the Skolem functions are obtained in some reasonable xed fashion. In this paper, we often suppress mention of the

294

revision:1996-03-11

CODING AND RESHAPING WHEN THERE ARE NO SHARPS

membership relation as a relation of a model, but it is always intended that it be one. Thus, (M, A, ) denotes the same model as (M, , A, ). All other notation is introduced as needed (we hope).

1. THE FORCINGS. (1.1) Denition. If g is a function, g = {x dom g : g(x) = 1}. (1.2) Denition. If is a innite cardinal, T , then g S(T, ) i theres = (g) (, + ) such that g : (, ) {0, 1} and for all (, ] : (),
g

(card )L[T,

g|]

= (we say: g promptly collapses ).

S(T, ) = (S(T, ), ). (1.3) Denition. Let be an innite cardinal, T + , g S(T, + ). b g = (bg : (+ , (g)]) is a sequence of almost disjoint conal subsets of successor ordinals (, + ) which are multiples of 3, such that for all (, (g)], (bg : (+ , ]) is canonically dened in L[T, g|].
modified:1996-03-11

(1.4) Denition. With , T, g as in (1.5), p = ( (p), r(p)) P,

T, g

revision:1996-03-11

(1) (p) S + (T , ), (2) if (, ( (p))), = 3 + 1, then (p)() = 1 i T . (we say: (p) codes T), + (3) r(p) : dom r(p) + , dom r(p) [dom g]< , and whenever dom r(p), r(p)() bg ( (p)), (p)() = g(), (4) if 1 , 2 dom r(p) and g(1 ) = g(2 ), then bg 1 \ r(p)(1 ) bg 2 \ r(p)(2 ) = . For p, q P, ).
T, g ,

p q i (p) (q), r(p) r(q); P,

T, g

= (P,

T, g ,

(1.5) Denition. Let be an innite cardinal. For n , let n be +n . Let Z be such that for all n , Hn = Ln [Z n ]. p S(, Z) = S() i dom p = , for all n < , p(n) = ( (p(n)), r(p(n))), (p(n)) S(Z n , n ) and p(n) Pn , Zn+1 , (p(n+1)) . For p, q S(), p q

294

SAHARON SHELAH1,2,5 AND LEE J. STANLEY3,4,5

i for all n < , (p(n)) (q(n)), r(p(n)) r(q(n)). S() = S(, Z) = (S(), ). If k < , Sk () = Sk (, Z) = {p|[k, ) : p S()}; k is the obvious projection of onto Sk (). Sk () = Sk (, Z) = (Sk (), k ).

2. THE RESULTS. Our ultimate goal in this section will be to prove that for cardinals with 2 , for all k < , Sk () is (k , )- distributive. As will be clear from what follows, by this we mean that the intersection k open dense sets is dense, and not the weaker notion involving fewer than k open dense sets. We denote the latter notion by (< k , -distributive. A useful rst step will be to establish something stronger than this latter notion. (2.1) Proposition. For all k < , Sk () is < k - complete. Proof. Let < k , (pi : i < ) be a k - increasing sequence from Sk (). For i < , k n < , let i (n) = ( (pi (n))), so, for such n, (i (n) : i < ) is non-decreasing. Let (n) = sup {i (n) : i < }. Let (p(n)) = { (pi (n)) : i < }, r(p(n)) = {r(pi (n)) : i < }, and let p(n) = ( (p(n)), r(p(n))), for k n < . We shall prove that p Sk (). The only diculty is to prove that for k n < , (card (n))L[Zn, (p(n))] = n . If is a successor ordinal or (n) = i (n) for some i < , this is clear. Otherwise, (n) is a limit ordinal of conality cf < n , so, by the Covering Lemma, already (cf (n))L < n . But then, since ( < (n))(card )L[Zn , (p(n))] n , the conclusion is clear. (2.2) Before proving the main lemma of the section, in (2.3), it will be helpful to simply remark (the proofs are easy, and the reader may consult [2] for an outline) that letting G be the canonical name for the generic, letting G(n) be the canonical name for { (p(n)) : p G}, and letting W (n) be the canonical name for G(n), then for all k < ,
Sk ()

revision:1996-03-11

modified:1996-03-11

(k n < )W (n), Z n L[Z k , W (k)] .

We shall use a variant of this fact with no further comment below, in the proof of the main lemma. We note only that by an easy density argument, it can be shown that for k n < and [n+1 , n+2 ), there is < n+1

294

CODING AND RESHAPING WHEN THERE ARE NO SHARPS

W (n+1)| such that whenever b \ , W (n)() = 0 W (n)( + 1) = (n+1)| W W (n + 1)(), and that { b : W (n)() = 0} is conal in n+1 . Thus, W (n + 1)() is read by: W (n + 1)() = i i there is a nal segment

x b

W (n+1)|

such that for all x, W (n)() = i.

(2.3) We are now ready for the main Lemma. Lemma. For all k < , Sk () is (k , )-distributive. Proof. We rst note that it suces to prove that for all k < ()k : Let p0 Sk (), let be regular >> 22 ; let < () be a wellordering of H in type ; let M = (H , < (), {Sk ()}, {Z}, {p0 }); let N M, k + 1 |N |, card |N | = k . Then there is p0 k p which is (N , Sk () |N |)-generic. The argument that ()k suces is well-known, so x the above data. Without loss of generality, we may assume that [|N |]<k |N |. It will often be convenient to work with the transitive collapse of N , so let : N N <k be the inverse of the transitive collapse map; thus, [|N |] |N |. Let = 1 = the transitive collapse map. If X |N | and (N , X) is amenable, then we let (X) = {(aX) : a |N |}, and similarly for (Y ) if (N , Y ) is amenable. We let n = (n ). We also let n = sup (|N |n ). For k < n < , note that n = (n )N , and that k+1 = k+1 . Note that by applying Proposition 2.1 to forcing over N with (Sk+1 ()), we easily construct p (Sk+1 ()) which is (N , (Sk+1 ()))-generic, such that (p0 )|[k + 1, ) is extended by p, in (k+1 ), such that for k + 1 | and all proper initial segments of p(n) lie in |N |. In n < , p(n) |N view of the discussion in (2.2), for forcing over N , p N [] |= (n)(k + 1 n < p(n) L[(Z k+1 ), p(k + 1)].

modified:1996-03-11

revision:1996-03-11

p Thus, N [] |= (Z ) L[(Z k+1 ), p(k + 1)]. A crucial observation is: Proposition. OR |N | < (( k+1 )+ )L . Proof. Let = OR |N |, = sup (OR |N |). Note that |LN : L 1 L , with critical point k+1 . If (( k+1 )+ )L , then 0 exists, which proves the Proposition. Thus, (cf n )L (cf k+1 )L , for all k + 1 n < . Typically, of course, k+1 is a (regular cardinal)L . Let xk+1 = Z k+1 , hk+1 = ((k + 1)). p

294

SAHARON SHELAH1,2,5 AND LEE J. STANLEY3,4,5

We shall construct in V, p(k) which is |N |-generic for Pk , xk+1 , hk+1 , as dened in N . Among other properties, letting hk = ((k)), hk will code p hk+1 . This will be clear from the construction; we shall use this fact before showing that (cf k+1 )L[Zk , hk ] = k . This is exactly what is required to show that if we dene p by letting p(n) = (( ((n))), (r((n)))) (recall p p , X) amenable), then p Sk () (and p our convention about (X) for (N is |N |-generic for Sk () |N |). We shall have hk = (qk ), r((k)) = r(qk ), where qi = ( (qi ), r(qi )) p and (qi : i k ) is dened recursively in V , with q0 = (p0 (k)). For this, in V , we let (Di : i < k ) enumerate the dense subsets, in |N |, of Pk , xk+1 , hk+1 , as dened in N . For all < k , (Di : i < ) |N |, in virtue of the closure property we have assumed for |N |. For all i < , well have qi |N |, so, by the same observation, for < k , (qi : i < ) |N |. Also, for j < k+1 , letting D(j) be the subset of Pk , xk+1 , hk+1 consist ing of those r with ( (r)) j, as dened in N , clearly D(j) is dense and so is among the Di . This will guarantee that sup {( (qi )) : i < k } = k+1 , provided that we know that qi+1 Di . This will be part of the construction and will also guarantee the genericity of p. For i < k , well set i = ( (qi )). For limit k , we let (q ) = { (qi ) : i < }, r(q ) = {r(qi ) : i < }. If < k , by the covering argument of the proposition of (2.1), these are always conditions, and, if < k , as noted above, (qi : i < ) |N |, so also q |N |. So, we must dene qi+1 , where our crucial work is done. For each i < < k+1 , a limit ordinal, we dene p, , 1 qi as follows: r(p, , 1 ) = r(qi ); if i < and 1 (mod 3) then (p, , 1 )() = 0 if Z & = 1, if Z, where is such that = 3 +1. If i +k , we x a subset b |N |L, b k which codes a wellordering of k in type , and for < k , we set (p, , 1 )(i + 3 + 2) = 0 if b & = 1 if b. If i + k < and 2 (mod 3), we set (p, , 1 )() = 0. Similarly, if < i + k , we set (p, , 1 )() = 0 for all i < such that 2 (mod 3). If i < and for some dom r(qi ), b k+1 \ r(qi )( ), then (p, , 1 )() = hk+1 ( ). Note that in virtue of (4) of (1.4), this is welldened. For all other successor ordinals, i which are multiples of 3, we set (p, , 1 )() = 0. Now, suppose is a limit ordinal, i < . We set (p, , 1 )() = 0, unless = & = 1), if = (in this case, we mark
h

294

revision:1996-03-11

modified:1996-03-11

CODING AND RESHAPING WHEN THERE ARE NO SHARPS

revision:1996-03-11

). Then, let p, , 2 p, , 1 be chosen canonically in Di . Now (, ) p, , 2 is denable in N , and so, for each , in N , we can compute a bound, () < k+1 , for sup {dom (p, , 2 ) : i < , a limit ordinal }, as a function of . Iterating in N gives us a club, Ei , of k+1 , Ei |N |. Now, (Hk+2 )N = Lk+2 [(Z) k+2 ], so all clubs of k+1 which lie in |N |, and, in particular, Ei , lie in L[(Z) k+2 ]. Already in L, card k+2 = card k+1 . So, in L[(Z) k+2 ] there is < ( k+1 )+ such that all clubs of k+1 which lie in |N |, in fact, lie in L [(Z) k+2 ]. This, L[(Z) k+2 ] however, readily gives us that unless (card k+1 ) = k (and in this case, there is no problem in proving that qk is a condition), there is a club C of k+1 , C L[(Z) k+2 ], such that C grows faster than any club of k+1 which lies in |N |. In particular, C grows faster than Ei , so that for suciently large < k+1 , all Ei -intervals above miss C. In V , x C C, o.t. C = k , C a club of k+1 . The idea of the above is that in constructing p, , 1 , we have marked and our hope is that in passing from p, , 1 to p, , 2 , we have not inadvertently marked anything else. While this is too much to hope for, in general, we shall be able to get that we have not marked anything else in C, provided we choose suciently large so that every interval of E i , above , misses C. So, GOODs winning strategy, nally, to go from qi to qi+1 , is to take to be the least ordinal > i , C which, as above, is suciently large that the interval [, ()) C = , and such that there is [i , ) C and then to take qi+1 = p, , 2 . Thus, GOOD has marked a member of C and nothing else in C, while obtaining qi+1 Di . Now, since, as remarked above, we know from the construction that hk codes hk+1 , in L[Z k , hk ], we can recover (Z) k+2 , and therefore C. But then, by the construction, we have that { C : (hk (), hk ( + 1)) = (1, 1)} is a conal subset of C . Thus, as required, (cf k+1 )L[Zk , hk ] = k . This completes the proof. (2.4) Taken together, (2.1) - (2.3) give us the following Lemma, which, in turn, gives us the Theorem of the Introduction: Lemma. Forcing with S() preserves conalities, GCH, and if G is V -generic for S(), then, in V [G] there is W + such that V [G] = V [W ], Z L[W, Z ] and for all n , Hn = Ln [W ] and for < < + , (card )L[W ] = .

294

modified:1996-03-11

10

SAHARON SHELAH1,2,5 AND LEE J. STANLEY3,4,5

Proof. Of course W = { (p(0)) : p G}. It is a routine generalization of arguments from Chapter 1 of [1] to see that for all k, there is Qk V Sk () such that S() Sk () Qk , and Sk () Qk is k+1 c.c. and card Qk = = k+1 . Further, for k = 0, (2.3) gives us that S() is (, )-distributive and clearly card S() = + . Thus, preservation of GCH is clear, as is the preservation of all cardinals except possibly + . The argument here S() is routine: if this failed, then letting = (cf + )V , for some 0 < k < + V Sk () , = k . But then, since (cf ) > k , forcing with Qk over V Sk () would have to collapse a cardinal k+1 which is impossible. (2.5) Remarks and Applications. (1) If we start from an arbitrary Z , we can, of course, code Z by rst coding Z into a Z, as above (e.g., by coding Z into Z on odd ordinals), and then proceeding as above. (2) In work in progress, we are attempting to develop a combinatorial approach to coding the universe by a real (when 0 does not exist). Part of our approach is to use the Easton product of the S(), for = 2 , or a limit cardinal, as a preliminary forcing, to simplify the universe before doing the main coding.
modified:1996-03-11

(3) Several people have observed that the S() aord a method of coding any set of ordinals using a set forcing over models of GCH where 0 does not exist. This can be done as follows. Let X , and assume, without loss of generality, that 2 . Code X into a Z + , where Z has the properties assumed above. Then, force with S() to get W , as above. Finally, since W reshapes the interval (, + ), we can continue to code W down to a real, using one of the usual methods of coding by a real. (4) Woodin has introduced the following abstract notion of condensation. A has condensation i theres an algebra, A V with underlying set , such that for any generic extension V of V : (*) if X and X is the underlying set of a subalgebra of A, and : (A , A ) (A|X, A X), where is the inverse of the transitive collapse map, then A V . has condensation i for all A , A has condensation. This

294

revision:1996-03-11

CODING AND RESHAPING WHEN THERE ARE NO SHARPS

11

notion has been investigated by Woodins student, D. Law, in his dissertation [3], and by Woodin himself. S. Friedman has observed that using (3), above, it can be shown that for any cardinal , we can force condensation for via a set forcing. We omit the proof, except to say that according to Friedman, this is not a routine consequence of the usual sort of condensation for L[r], but rather involves a closer look at the coding apparatus.

References
1. A. Beller, R. Jensen and P. Welch, Coding the Universe,, London Mathematical Society Lecture Notes Series, vol. 47, Cambridge University Press, Cambridge, 1982. 2. S. Friedman, A guide to Coding the universe by Beller, Jensen, Welch, J of Symbolic Logic 50 (1985), 1002- 1019. 3. D. Law, Doctoral dissertation, California Institute of Technology, to appear. Hebrew University & Rutgers University Lehigh University

294

revision:1996-03-11

modified:1996-03-11