3

Edit: In writing out more details, I realized I needed to make a slight modification to the question. I switched all proper containments from the old version of the question to "essential containment".

This question arises in looking for a counterexample to a claim made in the book by Dunford and Schwartz, but seems interesting in its own right. The question I really want to answer is: "Does there exist a Lebesgue measurable set $C \subseteq [0,1]$ with $m(C)>0$ s.t. for every Lebesgue measurable $m(S \cap C^c) = 0$ s.t. $m(S) >0$, the set $\{x : m((x+S) \cap C^c) =0\}$ is Lebesgue null?"

It would suffice to show the following claim which seems possibly true: "Let $C$ a fat Cantor set. Then for every $S \subseteq C$ s.t. $m(S)>0$, the set $\{x : m((x +S) \cap C^c)=0\}$ is countable."

Edit: Due to a request in the comments, I'll post a simplification of the claim made in Dunford and Schwartz that inspired the question. Let $(A, \mathcal{A},\alpha), (B, \mathcal{B},\beta)$ $\sigma$-finite measure spaces and $(A \times B, \mathcal{A} \otimes \mathcal{B}, \alpha \times \beta)$ their product. Let $f : A \times B \to \mathbb{R}$ measurable. Then there exists a sequence $\phi_n$ of finite linear combinations of indicators on measurable rectangles (sets of the form $S \times T, S \in \mathcal{A}, T \in \mathcal{B}$) s.t. $\phi_n \to f$ pointwise a.e. and $|\phi_n| \leq |f|$.

If we have a set $C$ of the above type, let $D := \{(x,x) + (c,0) : c \in C, 0 \leq x \leq 1\}.$ Then the above property shows that if $\phi$ is a finite linear combination of indicators of measurable rectangles s.t. $|\phi| \leq 1_D$, then $\phi = 0$ a.e. This contradicts the Dunford and Schwartz claim as we can show that $m(D)>0$.

  • @DonThousand Both claims, or just the one about the Cantor set? Could you provide justification, if you have it? – Physical Mathematics Jun 22 '21 at 15:45
  • @DonThousand I may be missing something obvious, but I don't immediately see how the linked post resolves the above question. – Physical Mathematics Jun 22 '21 at 15:48
  • I'd be interested in seeing the original claim (or point me to where it is in the book), assuming it's not just "there is no Lebesgue measurable set that does this thing" – Michael Jesurum Jun 22 '21 at 17:19
  • 1
    @MichaelJesurum The claim above is pretty far from the original claim in Dunford and Schwartz. I tried to sketch the motivation in an edit. – Physical Mathematics Jun 22 '21 at 17:38
  • Doesn't the following show that the claim in the book is false? https://math.stackexchange.com/questions/149968/a-set-of-positive-measure-contains-a-product-set-of-positive-measure See also here: https://math.stackexchange.com/questions/42748/measurable-rectangles-inside-a-non-null-set and here https://math.stackexchange.com/questions/3845826/does-every-positive-lebesgue-measure-set-in-mathbbr2-contain-a-product-of – PhoemueX Jun 23 '21 at 17:55
  • @PhoemueX Yes, it does. Thanks. The actual posted question is still interesting IMO, but that does resolve the motivating problem. – Physical Mathematics Jun 23 '21 at 19:03

0 Answers0