Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Given a closed bounded set $X \subset \mathbb{R}^3$ and two curves $\gamma_1$ and $\gamma_2$ in the group of orientation preserving isometries of $\mathbb{R}^3$. Define the sets $X_1$ and $X_2$ as the infinite intersections $X_1 = \bigcap_{p \in \gamma_1} pX$ and $X_2 = \bigcap_{q \in \gamma_2} qX$ where $kX$ represents the set $X$ transformed by $k$. If each of the $pX$ and $qX$ are measurable sets with measure $\mu$ defined on $\mathbb{R}^3$, and for each $p \in \gamma_1$ there exists a distinct $q \in \gamma_2$ such that $\mu(X \cap pX) \ge \mu(X \cap qX)$, can I conclude that $\mu(X_2) \le \mu(X_1)$ ? How so/ Why not? (We can assume the identity element of the isometry group is in both $\gamma_1$ and $\gamma_2$ if needed)

EDIT: Changed conclusion from $X_2 \subseteq X_1$ to $\mu(X_2) \le \mu(X_1)$ per Andreas Blass's comment.

share|improve this question
Did you mean to ask whether $\mu(X_2)\leq\mu(X_1)$? With the conclusion as stated, $X_2\subseteq X_1$, there are easy counterexamples. Let $X$ be the unit cube. Let $\gamma_1$ consist of leftward translations by distances ranging from 0 to $1/2$, and let $\gamma_2$ be similar but with rightward translations. –  Andreas Blass Apr 25 '11 at 15:08
Yes, thanks. I did mean to ask $\mu(X_2) \le \mu(X_1)$. –  Sai Apr 25 '11 at 15:20
add comment

1 Answer

up vote 1 down vote accepted

Here's a counterexample in $\mathbb{R}^2$; to get one in $\mathbb{R}^3$ just take the product with $[0,1]$. Let $X=[0,2]\times[0,1]$. Let $f:[0,1]\to[0,1]$ be a continuous function with $f(0)=0$, $f(1)=0$, $f(x)>0$ for $0<x<1$, and $(2-x)(1-f(x))$ strictly decreasing. (If my arithmetic is good, $f(x)=x-x^2$ works, but even if my arithmetic is bad, some $f$ works.) Let $\gamma_1$ consists of the translations by the vectors $(t,f(t))$ for $t\in[0,1]$, and let $\gamma_2$ consist of the translations by $(t,0)$. In both cases, the measure of the intersection of $X$ and its image under the $t$ translation begins at $\mu(X)=2$ for $t=0$ and decreases monotonically and continuously to $\mu([1,2]\times[0,1])=1$ at $t=1$. So the hypothesis in the question (that for each $p$ there is an appropriate $q$) is satisfied (in fact with equality of the two measures). But $X_2=[1,2]\times[0,1]$ while $X_1$ is a proper subset of that, missing some points near the bottom edge because $\gamma_1$ lifted the rectangle $X$ a little ways above the $x$-axis during the motion.

share|improve this answer
Thanks Andreas. I am not quite sure if I understand the example since the intersection of $X$ with the transformation $(1,1)$ produces a measure zero set. But I think I understand the rationale, the hypothesis requires one $q$ for every $p$ globally, and is a a bit weak. So, what happens if the hypothesis is refined to 'For each $p \in \gamma_1$ there exists a distinct $q \in \gamma_2$ in a neighborhood of $p$ such that $\mu(X \cap pX) \ge \mu(X \cap qX)$' (such a neighborhood can be constructed for example by picking $q$ in a ball around $p$ in $\gamma_1^c$) –  Sai Apr 25 '11 at 16:31
Thanks for catching a typo; $f(1)$ should be 0, not 1. ($x-x^2$ now works.) I edited the answer to correct this. As for requiring $q$ to be close to the corresponding $p$, I would expect that once you specify how close you want it, there will still be counterexamples similar to the one I described. Just take $f$ very small. –  Andreas Blass Apr 25 '11 at 16:50
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.