amsppt.stypsSFp90.tex Postscript to Shelah & Fremlin p90 D.H.Fremlin University of Essex, Colchester 25 September 1991 1.

The statement () of Shelah & Fremlin p90, 2G, can be strengthened, as follows. Write ( ) for the statement there is a closed negligible set Q [0, 1] such that (L ) (Q1 [D]) D for every D [0, 1]. L Proposition If P is a partially ordered set as in Shelah & Fremlin p90, then 1 P 1
P

( ).

proof (a) Take X and P and and as in Shelah & Fremlin p90, 1. Then if D X and > 0, there is a closed set F X, with F D , such that 1 P P F R 1 [D], 1 writing F for the P-name for a closed subset of X corresponding to F . P Choose k0 such that 2k0 1 , and a closed set F0 such that F0 = (D F0 ) D 1 . Set P 4 2 F = {x : k k0 , (D {w : w k = x k}) > 2k+1 {w : w k = x k}}. Then F is closed, and F0 \ F {x : x F0 , k k0 , (F0 {w : w k = x k}) 2k+1 {w : w k = x k}} has measure at most 2k0 +1 1 , so F D . 2 Now suppose that is a P-name such that 1P P F . 1 Set r = 1, Lk = nk for every k N and follow the argument of Lemma 1R of Shelah & Fremlin p90 down to the end of part (c), but insisting at the beginning that k0 k0 . Observe that p3 P s, where s = Hi (v ) i<k1 , as in part (d) of the proof of Lemma 1R. Also #(Jk ) 2k nk for all k k0 , so i that {x : s x, x(k) Jk k k1 } > (1 2k1 +1 ){x : s x}; / but as p3 P F , s , we must have (D {x : s x}) > 2k1 +1 {x : s x}, and there is a z D such that s z and z Jk for every k k1 . / Now continue the argument as in (e)-(i) of the proof of Lemma 1R to get p5 p3 such that p5 P (, z ) R . Q Q (b) Now we nd that 1 P P D X , ( R 1 [D]) D. 1 P Let 0 be a P-name for a subset of X, and , (ground-model) rationals such that P 1 P P 0 > > . 1 Take < such that whenever is a P -name for a closed subset of X and 1 P P > 1 1 then there is a P -name for a member of 0 . Now taking to be a P -name for the subset of X consisting of those members of 0 which can be represented by P -names, we see that 1 P P . 1 P Using part (a) in V , we have a P -name for a closed subset of X such that 1 1 P P & 1 P P () R 1 [] , 1 expressing P as an iteration P P as in the proof of Theorem 1S of Shelah & Fremlin p90. But this must mean that 1 P P ( R 1 [0 ]) , 1 and as was arbitrary this proves the claim. Q Q (c) This proves the result for (X, ) rather than for ([0, 1], L ). But as remarked in 2G of Shelah & Fremlin p90, we have a continuous inverse-measure-preserving f : X [0, 1] such that f 1 [D] = [D] L
1

406a

revision:1994-11-24

modified:1994-11-24

for every D [0, 1]; so that setting Q = {(f (x), f (y)) : (x, y) R} we obtain the result for Lebesgue measure, as stated. 2. In fact wse can go a little further: in V P , Q has the property that whenever D [0, 1] and E is a measurable set such that (E \ D) = 0, then (E \ Q1[D]) = 0. To see this, follow the arguments above, observing that it is enough to consider closed E, and that the set F of part (a) of the proof can be taken to be a subset of E. Reference Shelah S. & Fremlin D.H. [p90] Pointwise compact and stable sets of measurable functions, to appear in J. Symbolic Logic.

406a

revision:1994-11-24

modified:1994-11-24