saharon shelah- more on cardinal arithmetic

Upload: jtyhmf

Post on 06-Apr-2018

228 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    1/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC

    SH410

    Saharon Shelah

    Institute of MathematicsThe Hebrew University

    Jerusalem, Israel

    Rutgers University

    Mathematics DepartmentNew Brunswick, NJ USA

    Done 1990Partially supported by the Basic Research Fund, Israel Academy of Science

    I thank Alice Leonhardt for the beautiful typingPubl. 410Latest Revision - 07/July/3

    Typeset by AMS-TEX

    1

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    2/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    2 SAHARON SHELAH

    An Annotated Table of Contents

    1 On Normal Filters

    In [Sh 355, 4] and [Sh 400, 3,5] we have computed cov(,,,) when > =

    cf > 0, using tcfi

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    3/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 3

    6 The Existence of Strongly Almost Disjoint Families

    We characterize such existence questions by pps. An example is the question ofthe existence of a family of+ subsets of > 0 , each of cardinality (> 0) such

    that the intersection of any two is finite.

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    4/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    4 SAHARON SHELAH

    1 On Normal Filters

    The following Lemma 1.2 is similar to [Sh 355, 5.4], [Sh 400, 3.5], but deal with

    normal ideals (see [Sh 371, 4], in particular [Sh 371, Definition 4.1,Claim 4.6]).Remember prc is defined in [Sh 371, 4] as:

    1.1 Definition. 1) For a regular uncountable cardinal , normal ideal J on , a-sequence of cardinals > , and f Ord, we define:

    prcJ(f, ) = Min

    |P | :P is a family of-sequences of sets of ordinals,

    B = Bi : i < , |Bi| < i or at least

    {i < : |Bi| i} J, such that: for every g

    Ord,g J f there is a sequence A

    : < of members

    P satisfying {i < : g(i) / .

    () i strictly increasing and for limit i:

    if i =j

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    5/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 5

    1.3 Remark. 1) On getting =+ see [Sh 420, 6.1](C) and [Sh 430, 4]. The problemis when pcf(a) has an accumulation point which is inaccessible.2) In the case () holds, if = i : i < is (strictly) increasing continuous,

    supi

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    6/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    6 SAHARON SHELAH

    Proof of the Inequality from the Fact. Let us define a family P :

    P = {Bi, : < i and |Bi,| < i} : i < : Bi, : i < , < i P

    .Now each member ofP has the right form as each i is regular and Bi, : i we have B,i : i < P; let us enumerate

    > as { : < } such that [ { : < }], and let us define

    B

    = B,i : i < P

    , hence by Definition 1.1 it is enough to show thatE =: {i < : g(i) /

    the

    set X =: {i A : i /j]}

    X =: {i < : if i / Y then for some >i we have i X}

    belong to J. It suffices to show() for every \X, for some > we have B,i.

    Why () holds? Choose by induction on n < , n n such that: An . For

    n = 0 remember A = . For n + 1, as An and / X clearly / Xn so

    necessarily j

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    7/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 7

    for the given B and any g, i : i < as there. Let Yi0 = { < i : |Bi,| cf|B

    ,i|} and Y

    i2 = { < i : cf|B

    i,| i} (for

    each i < , < i).

    Clearly Yi0 , Yi1 , Yi2 is a partition of { : < i}, now for Yi1 Yi2 , leti =: cf |B

    ,i|, and Bi,, : <

    i be an increasing continuous sequence of

    subsets of Bi, of cardinality < |B,i| and

    sup( pcf(a) holds as is a

    successor cardinal.

    Case 2: Assume assumption. Here we partition a to sets diagonally; i.e. without

    loss of generality each i is a cardinal hence by clause () we have i j

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    8/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    8 SAHARON SHELAH

    Proof. For (c) see [Sh 371, 1.6], so we can assume 1 > 0 as otherwise we havethere gotten a conclusion stronger than (a) + (b) + (c).

    Let a Reg have cardinality < , be unbounded in , I a 2-complete

    ideal on a, and (

    < )[a

    I] and = tcf(a/I). As cf() = < < without loss of generality < Min(a), and let i : i < be increasing continuouswith limit , 0 cf(i) < (remember > 0); without loss of generality < 0 < Min)a) and pp(,1)(i) < i+1 and, if i > 0 is a limit ordinal thena i is unbounded in i. Also without loss of generality for every i < :

    ()1 0 < i & 1 cf() < pp(,1)() < i+1

    hence

    ()2 b (0, i) & |b| < sup pcf1-complete(b) < i+1.

    Let b[a] : pcf(a) be a generating sequence (exists by [Sh 371, 2.6]) andwithout loss of generality = max pcf(a). By [Sh 345a, 3.6] (and 3.1(7)) withoutloss of generality b[a] b[a] b[a]. Let i < satisfy cf(i) 1, as|a| < and ()1 we have sup pcf1-complete(i a) < i+1 hence for some ci we

    have: ci i+1 pcf(a i), |ci| < 1 and a i ci

    b[a] (otherwise we can

    find a 1-complete proper ideal J on i a such that

    [ < pp+(,1)(i) & pcf(i a) b[a] (i a\i) J],

    a contradiction to ()1. Note that ci i+1.Let S0 = { < : cf() 1}, so for some i() < we have:

    S1 = { S0 : c i()}.

    is a stationary subset of .

    By renaming, without loss of generality i() = 0 and so ci (i, i+1), and fori S1 let (i, , ei,) : < i list {(, b[a]) : ci}; so:

    ()1 a i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    9/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 9

    and by 1.6 pcf2-complete

    iS1

    di

    and by ()2 above, for j < implies

    / pcf1-complete iS1

    di j

    ; also = max pcf(d). So we are left with clauses

    (a) + (b). Let d = {, : < < 1}, as 1 < , clearly without loss of generalityfor some ()

    S3 = { S2 : = ()} is stationary.

    For each < () let

    P =: {S S2 : max pcf{, : S} < }.

    If for some < (), the normal ideal on which P generates is proper, we havefinished. If not, for each < () there are members S,i(i < ) ofP and clubC of such that:

    S3 C i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    10/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    10 SAHARON SHELAH

    2 On measures of the size of D cf() = 0, > 2 and [ < & cf() pp(

    ) < ] and pp() < cov(,, 1, 2). Then { : < = pp()} hasorder type .

    We shall return to this in [Sh 430, 1] so we do not elaborate.

    2.2 Claim. Suppose , are regular, 0 < < 1 , and ()[

    1 & cf () < pp

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    11/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 11

    (ii) the function b fb : < max pcf(b) (for b a) is as in [Sh 371, 1];i.e. satisfies:

    ()1 fb b, we stipulate fb (a\b) = 0a\b and f

    b : < is

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    12/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    12 SAHARON SHELAH

    Proof. Let = 3()+, and for + let B be an elementary submodel of(H(), ,

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    13/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 13

    (2) holds.5) Assume

    (a) (+

    )V

    = (+

    )W

    , (), ()+ ;(b) there is C : ( + 1)\( + 1), 0 cf

    W W satisfying C a clubof for each such that [ acc(C) C = C ] and otp(C) .[nec?]

    Then (W, V) satisfies the -strong covering (see [Sh:g, Ch.VII]) which means:

    for every ordinal and model M V with universe , with countablevocabulary, there is N M of power< , N an ordinal and the universeof N belongs to W.

    6) Moreover, in (5)

    + in the game where a play last moves, in the ith move (for i < ) the first

    and second players choose ai, bi []

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    14/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    14 SAHARON SHELAH

    There is no problem to carry the definition. Let N = N =:n

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    15/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 15

    by [Sh 400, 3.3A] we know M = N , so Y M M W andW |= |M | M < as required. [Saharon: fill details]

    Proof of Part (3). As in part (2), we choose () = V1 and get Ni, i, gi, hi(for i < ()). By 0 applied to {hi : i < ()} (again translating to a set ofordinals) there is a set B W such that W |= |B| < and A B infinite,without loss of generality B H (see proof of part (2)). So there is a limit ordinal < () such that = sup{i < : hi B}. Now let M be the Skolem Hull in

    (H()W, ,

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    16/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    16 SAHARON SHELAH

    As [i < j hi < hj ] clearly we get:

    () i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    17/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 17

    Now we shall show that for E, N W. For E we define by inductionon n < , Mn :

    M0 is the Skolem Hull in (H()

    W

    , , and for anya Reg we have: if |a| < , Min(a) > and a , even sup(a) < and I is a-complete ideal on a, then tcf(a/I) (when tcf is well defined).2) If we write instead of we mean for some , < .3) If we omit we mean = 0.4) accessible is just the negation of inaccessible.

    We now rephrase various old results.

    3.2 Claim. 1) For regular, in the definition, and -complete I, tcf(a/I)can be replaced by / pcf-complete(a) and also by = tcf(a/I) for any a Reg (, ), |a| < , I being -complete; also if cf() / [, ) then sup(a) <

    is not necessary just / a.2) Assume > > and cf() . Then is ( , ,)-inaccessible iff[ (, ) & cf() < pp(,)(

    ) < ].3) If = cf() > = cf() > , is ( , ,)-accessible then there is a seta Reg (, ) of ( , ,)-inaccessible cardinals each > , a of cardinality < such that pcf-complete(a).4) If = cf() > > = cf(), and (a)[a Reg (, ) & |a| < &

    pcf-complete(a)] then there is a set a of ( , ,)-inaccessible cardinals (, )with |a| < such that = max pcf(a), and pcf-complete(a). If is ( , ,)-

    inaccessible then necessarily sup(a) = , Jbda J

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    19/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 19

    (ii) for limit, = (

    (ii) for limit, = (

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    20/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    20 SAHARON SHELAH

    3.6 Claim. Assume |a| < Min(a), then

    cf|a|(pcf(a)) max pcf(a).

    Proof. More is proved in [Sh 371, 3].

    The following answers a question of Gerlits, Hajnal and Szentmiklossy in [GHS].They dealt with -good topological spaces X (i.e. every subset is the union of compact sets) and weakly -good spaces (every Y X of cardinality > contains a compact subset of cardinality > ). [GHS] has the easy implication.

    We return to this in [Sh 513, 6].

    3.7 Theorem. The following conditions on < < are equivalent: ( is aninfinite cardinal, and are ordinals)

    (A),, there are functions f : for < such that:

    < i 2, pp+() > 1.

    Proof. First note

    3.8 Observation. Let < , an infinite cardinal, , are ordinals. If for everyregular 1, 1 < the statement (A),,1 holds and is singular (e.g. > ||),then (A),, holds.

    Proof. We prove this by induction on ; if trivial: use the constant functions.

    As is singular =

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    21/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 21

    (B) (A)

    First Case: 2 ||.Let {A : < } be a family of || distinct subsets of , let A

    = {2i : i

    A} {2i + 1 : i / A, i < } and let f : {0, 1} be

    f(i) =

    0 if i A1 if i / A.

    Second Case: ,cf() , > 2, pp+() > , a regular cardinal.So there are regular cardinals i <

    for i < (such that i > ) and an ideal

    J on , / J such thati

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    22/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    22 SAHARON SHELAH

    hence gi j 2 (otherwise g() = sup(Ni g

    ())(as Ni (H(), , j(); asH(h) is not well defined, no < satisfies the requirements in (). But of the threedemands on , j trivially satisfy two and a half: < , f (\A) = g

    (\A)and f A g

    A; so the remaining one should fail, i.e. [h A fj A].So for some A we have h() > fj (); now h Njj()+1 Nj hence h() Njhence h() Nj \fj (), hence (by the definition of gj), gj() h() hence (as

    j S) we have g() h() but h G A

    g(), so h() < g(), contradiction.

    3.7

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    24/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    24 SAHARON SHELAH

    4 Entangled Orders Narrow Order Boolean Algebra Revisited

    4.1 Theorem. 1) If +4 cf() < 2 then there is an entangled linear orderof cardinality .2) Moreover, if 0 < we can demand that the linear order has density character 0 (in fact, in every interval of the linear order).

    Remark. See more [Sh 462] and [Sh 666].

    Proof. Without loss of generality 0 > cf() +4. By [Sh 355, 2.1] there is an

    increasing continuous sequence i : i < cf() of singular cardinals with limit

    such that tcf(

    i cf().

    Case I: For i < cf() we have max pcf{+j : j < i} < .

    So for some unbounded A cf() we have i A +i > max pcf{+j : j

    A i}.So a = {+i : i A} is as required in [Sh 355, 4.12] (with

    +, cf() here standing

    for , there, noting that 2cf() 2 ).

    So we can assume:

    Assumption not Case ISo there is , 0 < < , cf() < cf(), pp . Choose a minimal

    such , so by 3.2(2):

    () a Reg \0 & sup(a) < & |a| < cf() max pcf(a) < .

    Clearly (by [Sh 355, 2.3]) in ()s conclusion we can replace < by < i.e.

    () a Reg \0 & sup(a) < & |a| < cf() max pcf(a) < .

    Let =: cf(), so pp() = pp 0. As said above pp and by [Sh 371, 1.7]

    there is a strictly increasing sequence i : i < of regular cardinals satisfying

    =i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    25/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 25

    So +4 cf(). Let P =: {A : A cf(), otp(A) = +, A is a closed subsetof sup(A) and max pcf{+i : i c} =

    +sup(A)}. For any C P , try to choose by

    induction on i < +(bi) = bi[C] and i = i[C] such that:

    (i) bi Reg \j

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    26/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    26 SAHARON SHELAH

    i() > max pcf{i() : < },

    i() is strictly increasing, and Ens(

    i(),

    i()) for

    < (). So applying [\Sh:355] , 4.12] we finish.

    Why holds: Let i() < cf() be given. Choose C (i(), cf()) from P suchthat (b[C], [C]) is defined iff < [C] and [C] <

    +. By the definition ofPwe have max pcf{+ : C} < . Let d =: {

    + : C}, let b[d] : pcf(d)

    be as in [Sh 371, 2.6]. Let be minimal such that otp(b[d]) is . We canfind B C (for < ) such that {+ : B} b[d], otp|B| = and theBs are pairwise disjoint. Clearly max pcf{

    + : B} = as {

    + : B}

    is b[d], but is not a subset of any finite union of b [c], < . Now letting

    a =: {b[C] : < [C]}, there is (by [Sh 371, 2.6]) a subset a of a

    such that = max pcf(a) but / pcf(a\a). Now as pcf{+ : B},

    + pcf(b)

    we have (by [Sh 345a, 1.12]) pcf( B

    b) hence by the previous sentence

    pcf(a B

    b). Let c =: a B

    b , = , we can apply [Sh 355, 4.12] and

    get that there is an entangled linear order of cardinality (which is more thanrequired, see [Sh 345b]); and, of course, i() <

    pcf{j : j < cf()}. Theassumptions of [Sh 355, 4.12] holds as the c are pairwise disjoint (by (i) above),

    pcf{+ : B{ pcf(B

    b) = pcf(c) and [1 a max pcf(a 1) i() > 0.

    So holds and we finish Subcase IIIb hence Case III.

    Case IV: cf() > +4 and (and not Case I).For each < cf() of cofinality +4 we can apply the previous cases (or the

    induction hypothesis on ) and get an entangled linear order of power + . So holds and we finish as in Subcase IIIb. 4.1

    4.2 Claim. Assume +4 = cf(), i : i < is a strictly increasing sequence

    of regular cardinals, < i 2 and = tcf(

    i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    27/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 27

    Then (2

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    28/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    28 SAHARON SHELAH

    5 prd: Measuring f(i) by a family of idealsand family sequences Bi : i < , |Bi| < i

    In [Sh 371, 4], and here in 1 we have dealt with generalizations of the measuringi

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    29/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 29

    (iv) A A(v) f : A Ord

    (vi) if < > T and y A, f(y) = 0 then f(y) > f(y).

    5.5 Remark. 1) Clearly for 1 2 if is 2-suitable then is 1-suitable.2) If 2|Y| its value is immaterial, so we can omit it.

    5.6 Claim. 1) Y,,, if = cf() > 0 is closed, restriction closed, ofcharacter and is suitable. ForP P(Y),

    cY,,(P

    ) =

    Z : for some < , Ai P

    for i < we have Z i T).

    Let T =: { T : < g() () w}, as is regular uncountable, clearly|T| < . Let

    B =:{

    B

    :

    T

    } {A

    :

    T and A

    I} {

    f1

    ({

    0}

    ) :

    T

    }.

    Now B is the union of< members ofI and I is -complete, so B I. Choosey \B.

    We now choose by induction on n, n such that:

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    30/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    30 SAHARON SHELAH

    0 =

    n T,

    g(n) = n

    n n+1

    y An.

    For n = 0, 0 = so y Y = A. For n + 1, y Ann, n T and y / B

    hence y / Bn , so by () there is n wn such that y An so n+1 = nis as required.

    In the end for each n we have fn(y) > 0 as f1n ({0}) B, (remember n T

    )hence by condition (vi) from 5.4, fn(y) : n < is a strictly decreasing sequenceof ordinals, contradiction.2) Left to the reader.

    3) Again we leave the proof of restriction closed and closed and having character to the reader and prove suitability. The proof is similar but use diagonal union. Soby part (2) and condition (iii) of 5.4 for each T for some B J and functionh from to ordinals such that h(i) T for i < , we have

    () Y = B {\A0\A1\ . . . \Ag} {x Y : x

    i and each h the identity function.

    Let I be the normal ideal on Y generated by J {f1n ({0}) : T}, so we

    assume I is a normal proper ideal and we shall get a contradiction.Now define Y

    Y = {x Y : (a) (x) a limit ordinal <

    (b) if T >(x) then x / B

    (c) if T >(x) and A I then x / A}.

    ClearlyY

    Y

    mod I, hence we can find x() Y

    . Now we choose by inductionon nn >(x()) as in the proof of part (1) and get similar contradiction.

    4) Left to the reader. 5.6

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    31/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 31

    5.7 Definition. 1) For = i : i < , J an ideal on Y and f : Y ord wedefine

    prdJ(f, ) = Min

    |P| :P is a family of sequences of the form Bx : x Y,

    each Bx a set of ordinals, Bx of cardinality

    < (x) such that for every g xord satisfying

    g f, we have :

    P() = c[J {{x Y : g(x) Bx} : Bx : x Y P}]

    .

    2) If above J I, I an ideal on

    prdJ,I(f, ) = Min

    |P| :P is a family of sequences Bx : x Y,

    each Bx a set of ordinals of cardinality < (x)

    and for every g xord satisfying q f,

    we have :

    I c(J {{x Y : g(x) Bx} : Bx : x X P})

    .

    3) If is constantly , we may write instead. We can use also = x : x Y,(but usually do not) with the obvious meaning.

    5.8 Claim. 1) If {x Y : f(x) < } J we can in 5.7 demand g |Y|, {x : f(x) > } J then prdJ(f, ) cf().

    4) If f1 J f2 or just {x : |f1(x)| > |f2(x)|} J then prdJ(f1, ) prd

    J(f2, )

    (and the other obvious monotonicity properties).

    5) Ifi = , cf() > |Y| we can in Definition 5.7 demand

    xYBx = B0 forB P

    (i.e. without changing the value).6) If is -complete and restriction closed, () < , A : < () is a partitionofY (J, as in 5.7) then

    sup

    prdJ+(\A)(f, ) prdJ(f, )

    prdJ+(\A)(f, ).

    7) If is normal and restriction closed, A {x Y : (x) > } for < , A : < a partition ofY\1({0}) then

    sup

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    32/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    32 SAHARON SHELAH

    for eachi, cf(1i ) < 1i &

    21 = (

    1i )+. Then for (any , J and f YOrd) we

    have prdJ(f, 1) = prdJ(f,

    2).10) If = i : i < is increasing continuous with limit , and is normal,

    J then prdJ(, ) and even prd

    J(, ) = cf().

    11) Ifi |Y|, = Y,, then for any , prd{}(, ) =

    cov(||, , |Y|+, ).

    Proof. E.g.6) The first inequality should be clear. Also the second: assume it fails, let =

    prdJ+(\A)(f, ), , i > |Yi|, = i

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    33/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 33

    5.10 Conclusion. 1) Suppose , , , J are as in 5.9, f Y ord and {x : cf[f(x)] =f(x) (x)} J then

    + prdJ(f, ) = + sup{prdJ(g, ) : g

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    34/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    34 SAHARON SHELAH

    5.12 Definition. Let 1 cfJ(ax : x Y) be

    suptcf xY

    x/I : J I and x ax for x Y.

    5.13 Claim. = i : i < is non-decreasing, is a suitable restriction close

    family of ideals onY, J , f Y ord andxY

    f(x) (x) = cf (x).

    1) If prdJ(f, ) is regular, then for some ax : x Y we have:

    (i) ax Reg f(x)+\(x)

    (ii) | xY

    ax| < if () of 5.9(1) and |ax|+ < (x) when 5.9(2)s assumptions

    hold

    (iii) = 1 cfJ(ax : x Y).

    2) If is inaccessible, , -complete and [ < cov(, 0, , ) < ] then

    without loss of generality |xY

    ax| < 0.

    Proof. 1) Like the proof of 1.1.2) Straight.

    5.14 Claim. Assume the hypothesis of 5.9.If g YOrd, and each g(s) is an ordinal (x) and f(i) = g(i) and let

    = prdJ(g, ) +

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    35/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 35

    6 The Existence of Strongly Almost Disjoint Families

    See [Sh 355, 0] on the history of the subject.

    6.1 Theorem. Assume J is an ideal on , not the union of 0 members ofJ, > 0.Then

    T3J() = T2J() T

    1J() T

    4J()

    where

    T1J() = TJ() = sup{|F| :F is a family of functions from

    to such that for f = g fromF we have f =J g}.

    T2J() = sup : there are ni < for i < and regulari, >

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    36/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    36 SAHARON SHELAH

    2) The supremum in the definition of T1J() is always obtained.[Why? IfF1,F2 are as there, F1 maximal |F1| < |F2| then for every f F2 thereis gf F, etc.].

    Proof. T2J() = T3J().

    Trivially T3J() T2J(); for the other direction let appear in the sup defining

    T2J(), as exemplified by < i, : < ni >: i < ; as max pcf{i, : i A, < ni}is always regular, without loss of generality is regular.

    By [Sh 355, 1]; more elaborately, for some a a =: {i, : < ni, i < }we have [a = = max pcf(a)] and / pcf[a\a] and a = a : i < as in the definition of T

    3J(), let

    a = {i, : i < , < ni} and f : < be a sequence of members of a whichis and we shall get acontradiction, without loss of generality F M0 . Clearly for every f

    we have:{g F : g =J f} has cardinality

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    37/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 37

    (h) ifb an, max pcf(b) and |b| < thenfn b {Max{fc, b : < n} : n < , < max pcf(b) and

    pcf(an) and c {b[an] : pcf(an)}}

    where b fb : < max pcf(b) for b an is a definable function (in (H(), ,

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    38/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    38 SAHARON SHELAH

    Proof of 6.1A(1). T4J() T1J(). Let An , An An+1, =

    n 0, > 0 and I = [] . ThenT4I() is T

    2I() hence is T

    1J().

    Proof. Apply 6.1 (+ here corresponds to there), more exactly by 6.1(A)(2).

    6.4 Remark. Asking on almost disjoint sets is an inessential change.

    Assignments

    1) Improve 1 ( less inaccessible by ), 0 < cf = < , for < large enough pp()() <

    +().2) Introduction on 3 and on 4.3) cf() = 1, = < 2

    1 ! No inaccessible below ? Or first such fix point?

    Similarly below 3(1) for more complicated cases.4) |a| < Min(a), acc[pcfa] has inaccessible member?5) Check that trivial computations on pp appear.6) Localization for a property ??

    > scite{4.5} undefined7) sup{|F| : f : , ({i : f(i) = g(i)}) < } for < ; appeared?8) has small stationary set = +, < , find for .9) In V1 we get -like demands for every > 3(1) from there to (old end of4)?.

    > scite{4.5} undefined

    10) Bound on pcf(a); or

    11) Better bound of when i : i < is increasing continuous +i pcf(a) (like

    the one of 4.4 rather than 2.1, 2.2.12) strong limit of cofinality 1, below it 1 inaccessible, (bound better than< (21 )+ inaccessibles.13) If |pcf(a)| |a|+4 then we have large fD for some f : |a|

    +4 ordinals, Dimplicit in 2.1, 2.2.14) Localization for -complete ([Sh 345]).15) = first (of cofinality 0 or 1) strong limit. Let pp() =

    +, fix point oflarge cofinality. Say 1 2 if (x (1 < 2 <

    + and (R [1, 1]) < ).16) pcf(a) large implies large rank.17) If =

    , < < pp() then (R ) is -directed mod {A R :

    for every a A, |a| cf we have max pcf A < } at least if in [, ] there is noinaccessible.

    May 10, 1992

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    39/40

    (410)

    revision:2007-07-

    03

    modified:2007-0

    7-04

    MORE ON CARDINAL ARITHMETIC SH410 39

    18) Proof of 1.6 say more?19) Quotations of [Sh 400, 3.3a,5.1](A)(1) accurate? I make distinction.20) Where is prdp(f, ) prd

    21) Note: more on proof of 2.2, 2.4?

    Revised Dec 2001:

    Compare 3-6, (1-2 earlier)

    4 - read, 5.1-5.6

  • 8/3/2019 Saharon Shelah- More on Cardinal Arithmetic

    40/40

    revision:2007-07-

    03

    modified:2007-0

    7-04

    40 SAHARON SHELAH

    REFERENCES.

    [GHS] J. Gerlits, Andras Hajnal, and Z. Szentmiklossy. On the cardinality ofcertain Hausdorff spaces. Discrete Mathematics, 108:3135, 1992. Topo-logical, algebraical and combinatorial structures. Froliks Memorial Vol-ume.

    [Sh:b] Saharon Shelah. Proper forcing, volume 940 of Lecture Notes in Mathe-matics. Springer-Verlag, Berlin-New York, xxix+496 pp, 1982.

    [Sh 345] Saharon Shelah. Products of regular cardinals and cardinal invariants ofproducts of Boolean algebras. Israel Journal of Mathematics, 70:129187, 1990.

    [Sh 420] Saharon Shelah. Advances in Cardinal Arithmetic. In Finite and Infi-

    nite Combinatorics in Sets and Logic, pages 355383. Kluwer AcademicPublishers, 1993. N.W. Sauer et al (eds.). 0708.1979.

    [Sh 371] Saharon Shelah. Advanced: cofinalities of small reduced products. InCardinal Arithmetic, volume 29 of Oxford Logic Guides, chapter VIII.Oxford University Press, 1994.

    [Sh 355] Saharon Shelah. +1 has a Jonsson Algebra. In Cardinal Arithmetic,volume 29 of Oxford Logic Guides, chapter II. Oxford University Press,1994.

    [Sh 345a] Saharon Shelah. Basic: Cofinalities of small reduced products. In Car-

    dinal Arithmetic, volume 29 of Oxford Logic Guides, chapter I. OxfordUniversity Press, 1994.

    [Sh 400] Saharon Shelah. Cardinal Arithmetic. In Cardinal Arithmetic, volume 29ofOxford Logic Guides, chapter IX. Oxford University Press, 1994. Note:See also [Sh400a] below.

    [Sh:g] Saharon Shelah. Cardinal Arithmetic, volume 29 ofOxford Logic Guides.Oxford University Press, 1994.

    [Sh 345b] Saharon Shelah. Entangled Orders and Narrow Boolean Algebras. InCardinal Arithmetic, volume 29 of Oxford Logic Guides. Oxford Univer-

    sity Press, 1994. Appendix 2.

    [Sh 430] Saharon Shelah. Further cardinal arithmetic. Israel Journal of Mathe-matics, 95:61114, 1996. math.LO/9610226.

    [Sh 462] Saharon Shelah. -entangled linear orders and narrowness of productsof Boolean algebras. Fundamenta Mathematicae, 153:199275, 1997.math.LO/9609216.

    [Sh 666] Saharon Shelah. On what I do not understand (and have some-thing to say:) Part I. Fundamenta Mathematicae, 166:182, 2000.math.LO/9906113.