Set of $x\in[0,1]$ such that 4 shows up in decimal expansion infinity often

71 Views Asked by At

I feel like I made a mistake, is this correct?

If $$E=\{\sum_{i=1}^\infty\frac{a_i}{10^i}\in[0,1]: a_i\in\{0,1,2,\dots, 9\} \text{ and } \forall n>0,\exists m>n, a_m=4\}$$ then define $$E_n=\{\sum_{i=1}^\infty\frac{a_i}{10^i}\in[0,1]: a_n=4\}.$$ Now, $x=\sum_{i=1}^\infty\frac{a_i}{10^i}$ is such that $a_n=4$ if and only if $$4/10^n\leq x\leq 5/10^n.$$ This gives us the measure of $E_n$.

We have that $E=\bigcap_{N=1}^\infty\bigcup_{k=N}^{\infty}E_k$ as the set on the right is equivalent to $$\{x:x\in E_k \text{ for infinitely many $k$}\}.$$

If this is true then, $$m(E)=\lim m\bigcup_{k=n}^\infty E_k\leq\lim \sum_{k=n}^\infty m E_k=0.$$

$\bf{Correction}$: $x\in E_n$ then the first $n-1$ digits can be any number $0-9$ and this gives $10^{n-1}$ possiblities. Then $y+\frac{4}{10^n}\leq x\leq y+\frac{5}{10^n}$ for some $y=\sum^{n-1}_{i=1}\beta_{i}/10^{i}$. Thus there are $10^{n-1}$ intervals of length $10^{-n}$.

Then $\sum mE_n=\infty$.

Fix $m,k$ and consider $E_m\cap E_{m+k}$. There are $10^{m+k-1}$ intervals that make up $E_{m+k}$. Each interval is determined by the proceeding $m+k-1$ digits of the left endpoint, by what we saw above. Since the $m$th digit is $4$, there are $10^{m+k-2}$ possible intervals in total, each of length $10^{-m-k}$. This argument leads to showing that $\{E_n\}$ is independent.

Now, Borel-Cantelli asserts that in a probability space if we are given a sequence of events, and if both the sum of these events diverges and the events are pairwise independent, then the lim sup has probability 1.

1

There are 1 best solutions below

0
On BEST ANSWER

The mistake is in thinking that $x \in E_n$ iff $\frac 4 {10^{n}} \leq x \leq \frac 5 {10^{n}}$. In fact $E_n$ is a union of $10^{n-1}$ disjoint intervals of length $\frac 1 {10^{n}}$ each, so $m(E_n)=\frac 1 {10}$. For each $n$. Using Borel - Cantelli Lemmas we can show that $m(E)=1$.