# Is measure preserving function almost surjective?

Let $F:[0,1]\to[0,1]$ be a Lebesgue measure preserving function. Is $F$ almost surjective, i.e., the image of $F$ has interior measure one?

This question is motivated by the following observation. If $F$ satisfies the above and $X$ is uniformly distributed over $[0,1]$, then $F(X)\sim Unif[0,1]$, i.e., $F(X)$ seems to appear almost everywhere in [0,1]. I believe the answer is no, but a counterexample is nontrivial.

A relating post is https://math.stackexchange.com/questions/2612075/is-every-measure-preserving-function-almost-surjective but the counter example therein does not apply to my question. (The function constructed there has different domain and range.)

Yes, by Luzin's theorem. Fix $$\varepsilon>0$$ and take a compact subset $$K$$ of measure at least $$1-\varepsilon$$ such that $$F$$ is continuous on $$K$$. Then $$F(K)$$ is a compact set of at least the same measure as $$K$$, since $$F^{-1}(F(K))\supset K$$. So, for any $$\varepsilon>0$$, $$F([0,1])$$ contains a measurable subset of measure $$\geqslant 1-\varepsilon$$. This implies that inner measure of $$F([0,1])$$ is 1.

• How does that answer the OP's question? Could you elaborate? – Algernon Mar 31 '18 at 15:30
• Dear Fedor: I was surprised that my guess was wrong, and your answer is great! A direct consequence of your answer is that $F([0,1])$ is measurable with measure one, since the outer measure of it has to be less than or equal to one. – Zuofeng Shang Mar 31 '18 at 18:00

A few remarks (where $m$ is the Lebesgue measure):

1. If $F[0,1]$ is a Lebesgue measurable set, then $F[0,1]$ has Lebesgue measure equal to one (you can read Fedor Petrov's comment for this post).
2. Because of Birkhoff's Ergodic Theorem: For every $[a,b]\subset [0,1]$ $$\int_0^1\left(\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n} 1_{[a,b]}(F^i (x))\right)dm=m([a,b]).$$
3. If $F[0,1]$ is a Lebesgue measurable set and $F$ is also Ergodic, then for $m$-a.e. $x$ $$1=\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n} 1_{F[0,1]}(F^i (x))=m(F[0,1]),$$ which is another proof that $m(F[0,1])=1.$
4. If $F[0,1]$ is a Lebesgue measurable set, then $$1=\int_0^1\left(\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n} 1_{F[0,1]}(F^i (x))\right)dm=m(F[0,1]),$$ which is another proof that $m(F[0,1])=1.$
5. If $T$ is ergodic and $K\subset [0,1]\setminus F[0,1]$ is a Lebesgue measurable set, then $m(K)=0.$ By contradiction, we would obtain that for $m$-a.e. $x$ $$0=\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n} 1_{K}(F^i (x))=\mathcal{Leb}(K)\neq 0.$$
6. If $K\subset [0,1]\setminus F[0,1]$ is a Lebesgue measurable set, then $m(K)=0.$ By contradiction, we would obtain that $$0=\int_0^1\left(\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n} 1_{K}(F^i (x))\right)d m=m(K)\neq 0.$$
7. Let $\mathcal{A}=\{K:K\subset F[0,1], K \mbox{ measurable }\}$ and $\mathcal{B}=\{K:K\subset [0,1]\setminus F[0,1], K \mbox{ measurable }\}.$ Then $\sup_{K\in \mathcal{A}} m(K)=1.$ By contradiction, if not, by sigma aditivity $$1=m[0,1]=\sup_{K\in \mathcal{A}} m(K)+\sup_{K\in \mathcal{B}} m(K)=\sup_{K\in \mathcal{A}} m(K)<1,$$ which is another proof that the image of $F$ has inner measure one.
• The problem is that $F[0,1]$ is not a priori measurable – Fedor Petrov Apr 18 '18 at 12:48
• Thanks Fedor, I have added the hypothesis to my answer. – user39115 Apr 18 '18 at 12:59
• Under this hypothesis there is nothing to prove: if $A=[0, 1]\setminus F[0,1]$, then the preimage of $A$ is empty, but it must have the same measure as $A$. By the way, ergodic theorem holds only for ergodic maps, not all measure preserving maps. – Fedor Petrov Apr 18 '18 at 13:55