Sie sind auf Seite 1von 9

STRONG POLARIZED RELATIONS FOR THE CONTINUUM

SHIMON GARTI AND SAHARON SHELAH ` ` 1,1 Abstract. We prove that the strong polarized relation 2 2 2 is consistent with ZFC. We show this for = 0 , and for every supercompact cardinal . We also characterize the polarized relation below the splitting number.

revision:2010-10-26

modified:2010-11-08

2000 Mathematics Subject Classication. 03E05, 03E55, 03E35. Key words and phrases. Partition calculus, forcing, large cardinals. First typed: February 2010 Research supported by the United States-Israel Binational Science Foundation. Publication 964 of the second author.
1

964

SHIMON GARTI AND SAHARON SHELAH

0. introduction The balanced polarized relation 2 asserts that for every color ing c : 2 there are A and B such that otp(A) = , otp(B) = and c (A B) is constant. This relation was rst introduced in [2], and investigated further in [1]. A wonderful summary of the basic facts for this relation, appears in [7]. Apparently, this relation can be true only when and . It means 1,1 that the strongest form of it is the case of . We can give a name 2 to this situation: Denition 0.1. The strong polarized relation. 1,1 If then we say that the pair (, ) satises the strong polarized relation with colors.
If 2 = + then 2 . This result, and similar negative results, go back to Sierpinsky. Despite the negative results under the (local) assump+ +

1,1

1,1

tion of the GCH, we can show that the positive relation 2 is consistent with ZFC. Such a result appears in [3], and it was known for the specic case of = 0 (under the appropriate assumption, e.g., MA + 20 > 1 , see Laver in [4] which proves that if Martins Axiom holds for 1,1 then 2 ). So the negative result under the GCH cannot become a theorem in ZFC. 2 1,1 But we can restate this result in the form 2 2 . In this light, a natural question is whether the positive relation 2 2 2 is consistent with ZFC. One might suspect that the answer is negative, and this is the correct generalization of the negative result under the GCH. Notice that the result of Laver (from [4]) does not help in this case, since MA20 never holds. Moreover, in the rst section we prove that if 0 < cf() < s, then 1,1 2 . Hence, for every below the continuum (with uncountable conality), we can force a positive result by increasing s. But s 20 , so this method does not help in our question. Again, under some assumptions one can prove negative results. For example, if 2 is regular and MA(countable) 2 1,1 holds, then 2 2 . Nevertheless, we shall prove that a positive relation is consistent here. In the rst section we deal with = 0 . Here we can use a nite support iteration of ccc forcing notions, yielding a ccc forcing notion in the limit of the sequence. We indicate that cf(2 ) 2 in our construction, and 2 1,1 it might be that the relation 2 2 is provable in ZFC whenever cf(2 ) = 1 . Notice that cf(2 ) = 1 implies the existence of weak diamond on 1 , so the negative relation becomes plausible. In the second section we try to generalize it to higher cardinals. Here we encounter some diculty, since the chain condition is not inherited from the members of the iteration. Starting with Laver-indestructible supercompact

1,1

1,1

964

revision:2010-10-26

modified:2010-11-08

STRONG POLARIZED RELATIONS FOR THE CONTINUUM

cardinal, we can overcome this problem (as well as other obstacles in the generalization of the countable case). We try to use standard notation. We use the letters , , , , for innite cardinals, and , , , , , for ordinals. For a regular cardinal we denote bd the ideal of bounded subsets of by J . For A, B we say that A B when A \ B is bounded in . The symbol [] means the collection of all the subsets of of cardinality . We denote the continuum by c. Recall that by Laver (in [5]) we can make a supercompact cardinal indestructible, upon forcing with -closed forcing notions. We shall use this assumption in 2. We indicate that p q means (in this paper) that q gives more information than p in forcing notions. An important specic forcing to be mentioned here is the Mathias forcing. Let D be a nonprincipal ultralter on . We dene MD as follows. The conditions in MD are pairs of the form (s, A) when s []< and A D. For the order, (s1 , A1 ) (s2 , A2 ) i s1 s2 , A1 A2 and s2 \ s1 A1 . Let G MD be generic over V. The Mathias real xG is {s : A D, (s, A) G}. Notice that in VMD we have (xG B) (xG \ B) for every B [] . We shall use this profound property while trying to create the monochromatic subsets in the theorems below. In 1 we use the original Mathias forcing, and in 2 we use the straightforward generalization of it for higher cardinals. We thank the referees for many helpful comments, corrections and improvements.

964

revision:2010-10-26

modified:2010-11-08

SHIMON GARTI AND SAHARON SHELAH

1. The countable case


c c We prove, in this section, the consistency of n , for every natural number n. The general pattern of the proof will be used also in the next section. 1,1

Theorem 1.1. A positive realtion for c. c c 1,1 The strong relation n is consistent with ZFC, for every natural number n. Proof. Choose an uncountable cardinal so that 0 = and cf() 2 . We dene a nite support iteration Pi , Qj : i 1 , j < of ccc forcing notions, such that |Pi | = for every i 1 . Let Q0 be (a name of) a forcing notion which adds reals (e.g., Cohen forcing). For every j < 1 let D j be a name of a nonprincipal ultralter on . Let Q1+j be a ccc forcing notion which adds an innite set Aj such B A \ B). The Mathias forcing M that (B D j )(Aj j D j can end, set P = {P : i < }. serve. At the i 1 Since every component satises the ccc, and we use nite support iteration, P is also a ccc forcing notion and hence no cardinal is collapsed in VP . In addition, notice that 20 = after forcing with P. Our goal is to prove 1,1 that n in VP . Let c be a name of a function from into n. For every < 20 we have a name (in V) to the restriction c ({} ). P is ccc, hence the color of every pair of the form (, n) is determined by an antichain which includes at most 0 conditions. Since we have to decide the color of 0 -many pairs in c ({} ), and the length of P is 1 , we know that c ({} ) is a name in Pi() for some i() < 1 . For every j < 1 let Uj be the set { < : i() j}. Recall that 1 < 2 cf(), hence for some j < 1 we have Uj [] . Choose such j, and denote Uj by U. We shall try to show that U can serve (after some shrinking) as the rst coordinate in the monochromatic subset. Choose a generic subset G P, and denote Aj [G] by A. For each U we know that c ({} A) is constant, except a possible mistake over a A. But this mistake can be amended. nite subset of For every U choose k() and m() < n so that ( A)[ k() c[G](, ) = m()]. n is nite and cf() > 0 , so one can x some k and a color m < n such that for some U1 [U] we have U1 k() = k m() = m. Let B be A \ k, so B [] . By the fact that U1 U we know that c(, ) = m for every U1 and B, yielding the positive relation c c 1,1 n , as required.
1.1

964

revision:2010-10-26

modified:2010-11-08

STRONG POLARIZED RELATIONS FOR THE CONTINUUM

Remark 1.2. Assume is an uncountable regular cardinal. Denote by 1,1 st n the assertion that for every coloring c : n there exists B [] and a stationary subset U so that c U B is constant. c 1,1 c Our proof gives the consistency of st n , when the continuum is regular. Let us turn to cardinals below the continuum. We deal with the relation 1,1 2 , when < s. We shall prove that this relation holds i is of uncountable conality. Recall:

Denition 1.3. The splitting number. Let F = {S : < } be a family of subsets of , and B [] . () F splits B if |B S | = |B \ S | = 0 for some < ( ) F is a splitting family if F splits B for every B [] ( )s = min{|F| : F is a splitting family} Claim 1.4. The polarized relation below s. Assume < s. 1,1 Then 2 i cf() > 0 . Proof. Suppose rst that cf() > 0 . Let c : 2 be a coloring. Dene S = {n : c(, n) = 0} for every < , and F = Fc = {S : < }. Since < s, we know that F is not a splitting family. Choose an evidence, i.e., B [] which is not splitted by F. It means that (B S ) (B \ S ) for every < . At least one of this two opstions occurs -many times, and since all we need for the rst coordinate (in the monochromatic subset) is its cardinality, we shall assume (without loss of generality) that B S for every < . For every < there exists a nite set t such that B \ t S . There are countably-many t -s, and cf() > 0 , so for some t []< and H0 [] we have H0 B \ t S . Set H1 = B \ t, and verify that c H0 H1 0, so we are done. Now suppose cf() = 0 , and choose an increasing sequence n : n which tends to . For every < , let () be the rst natural number n such that n < n+1 . Dene c(, n) = 0 () n. We claim that c 1,1 is an evidence to the negative assertion 2 . Indeed, assume H [] and B [] . If c H B 1 then () < n for every H and n B. Choose some specic n B. Since H is unbounded in , one can pick large enough H such that ( ) n. Consequently, c( , n) = 0, a contradiction. On the other hand, if c H B 0 then () n for every H and n B. Choose some specic H. Since B is unbounded in , one can pick large enough n B such that () < n . Consequently, c(, n ) = 1, a contradiction. So the proof is complete.
1.4

964

revision:2010-10-26

modified:2010-11-08

SHIMON GARTI AND SAHARON SHELAH 1,1

s s We indicate that for s itself we believe that the relation 2 is independent of ZFC. We hope to shed light on this issue in a subsequent work.

964

revision:2010-10-26

modified:2010-11-08

STRONG POLARIZED RELATIONS FOR THE CONTINUUM

2. The supercompact case In this section we prove the consistency of 2 2 2 for every supercompact . We shall use [5] for making indestructible (in fact, all we need is the measurability of at every stage of the iteration), and [6] for preserving the property of -cc along the iteration. We shall use a generalization of the Mathias forcing, so we need the following: Denition 2.1. The generalized Mathias forcing. Let be a supercompact (or even just measurable) cardinal, and D a nonprincipal -complete ultralter on . The forcing notion M consists of D pairs (a, A) such that a []< , A D. For the order, (a1 , A1 ) (a2 , A2 ) i a1 a2 , A1 A2 and a2 \ a1 A1 . If M is a -Mathias forcing, then for dening the Mathias -real we D take a generic G M , and dene xG = {a : (A D)((a, A) G)}. D As in the original Mathias forcing, xG is endowed with the property xG A xG \ A for every A [] . Theorem 2.2. A positive relation for 2 . 1,1 is consistent with ZFC, for every < . The strong relation 2 2

1,1

964

revision:2010-10-26

Proof. Let be a supercompact cardinal. Starting with Lavers forcing, we may assume that is Laver-indestructible. Choose any so that cf() = ++ , and = . Let P , Q : + , < + be an iteration with (< ) support. We start with P0 = {} and Q0 a name in P0 of a forcing which to (e.g., Cohen forcing). increases 2 For every < + we choose D , a name of a nonprincipal -complete name of) the generalized Mathias forcing ultralter on . Let Q1+ be (a is -closed (since D is a -complete ultralter), so MD . Notice that MD remains supercompact and hence measurable along the iteration. For every 0 < < + , choose a generic set G1+ Q1+ , and let A1+ be the Mathias -real associated with it. We shall work in VP , when 1,1 P = {P : < + }, aiming to show the positive relation 2 2 . First, let us indicate that P satises the + -cc. It follows from [6], upon noticing that each component satises a strong form of the + -cc as required there. Second, P is -complete, since each component is -complete. Consequently, no cardinal is collapsed and no conality is changed by P. Moreover, Q0 blows 2 to , and the completeness of the other forcing notions ensures 1,1 that VP |= 2 = . We shall prove that in VP . Assume that < is xed, and c is a name of a coloring function from into . We denote c ({} ) by c , for every < , and we claim that c P() for some () < + . For this, notice that {} consists of and for the color of each pair we have at most conditions which pairs,

modified:2010-11-08

SHIMON GARTI AND SAHARON SHELAH

give dierent values, since P is + -cc. But P is of length + , so c appears at some early stage P() . For every < + , set U = { < : () }. Since cf() = ++ > + , one can pick an ordinal < + so that |U | = . Let U be U , and let G P be generic over V. Denote A [G] by A. For every U, c is constant on A, except a small (i.e., of cardinality less than ) subset of A. For each U choose () and () such that () A c[G](, ) = (). By the assumptions on , and the conality of , there U1 [U] and , such that () is and () for every U1 . Set B = A \ , and notice that |B| = . By the above considerations, 1,1 for every U1 and every B we have c(, ) = . Hence , and the proof is complete.
2.2 Remark 2.3. Notice that the relation under GCH, as follows from [7] theorem 4.2.8. 1,1 2

is completely determined

964

revision:2010-10-26

modified:2010-11-08

STRONG POLARIZED RELATIONS FOR THE CONTINUUM

References
1. P. Erds, A. Hajnal, and R. Rado, Partition relations for cardinal numbers, Acta Math. o Acad. Sci. Hungar. 16 (1965), 93196. MR MR0202613 (34 #2475) 2. P. Erds and R. Rado, A partition calculus in set theory, Bull. Amer. Math. Soc. 62 o (1956), 427489. MR MR0081864 (18,458a) 3. Shimon Garti and Saharon Shelah, A strong polarized relation, in preperation. 4. R. Laver, Partition relations for uncountable cardinals 20 , Innite and nite sets (Colloq., Keszthely, 1973; dedicated to P. Erds on his 60th birthday), Vol. II, Northo Holland, Amsterdam, 1975, pp. 10291042. Colloq. Math. Soc. Jans Bolyai, Vol. 10. o MR MR0371652 (51 #7870) 5. Richard Laver, Making the supercompactness of indestructible under -directed closed forcing, Israel J. Math. 29 (1978), no. 4, 385388. MR MR0472529 (57 #12226) 6. S. Shelah, A weak generalization of MA to higher cardinals, Israel J. Math. 30 (1978), no. 4, 297306. MR MR0505492 (58 #21606) 7. Neil H. Williams, Combinatorial set theory, studies in logic and the foundations of mathematics, vol. 91, North-Holland publishing company, Amsterdam, New York, Oxford, 1977. Institute of Mathematics The Hebrew University of Jerusalem Jerusalem 91904, Israel E-mail address: shimon.garty@mail.huji.ac.il Institute of Mathematics The Hebrew University of Jerusalem Jerusalem 91904, Israel and Department of Mathematics Rutgers University New Brunswick, NJ 08854, USA E-mail address: shelah@math.huji.ac.il URL: http://www.math.rutgers.edu/~shelah

964

revision:2010-10-26

modified:2010-11-08

Das könnte Ihnen auch gefallen