4

Problem from Folland : based on Lebesgue Density Theorem: Let $D_{E}(x) = \lim_{r\to 0}\frac{\mu(E\cap B(r,x))}{\mu(B(r,x))}$ whenever it exists. Find examples of $E$ and $x$ such that $D_{E}(x)$ is a given number $\alpha \in (0,1)$ , or such that $D_{E}(x)$ does not exist. ($X = \mathbb{R}^n$,$\mu$ is Lebesgue measure)

user24367
  • 1,286
  • 3
    +5 for a problem copied from a textbook? Really? http://meta.math.stackexchange.com/questions/1803/how-to-ask-a-homework-question – Nate Eldredge Nov 01 '11 at 03:24
  • 1
    a problem is a problem!! doesnt matter if its textbook...!! – user24367 Nov 01 '11 at 09:16
  • 2
    As Nate is probably implying or suggesting, you need to show more indications of where you are stuck, which parts of the question you do or don't understand, and so on –  Jan 25 '12 at 04:53
  • Why does this question has 6 negative votes? There are a lot of questions more stupid with a lot more positive votes... http://math.stackexchange.com/questions/54506/is-this-batman-equation-for-real – Beni Bogosel Apr 03 '13 at 07:25

2 Answers2

10

For the second part, let $B_n=B(1/n,0)=(-1/n,1/n)\subseteq\mathbb{R}$ for every $n\in\{1,2,\ldots\}$ and $$E=\bigcup_{n=1}^\infty(B_{(2n-1)!}\setminus B_{(2n)!}).$$

If $n$ is odd, then $B_{n!}\setminus B_{(n+1)!}\subseteq E$. Hence $$\frac{\mu(E\cap B_{n!})}{\mu(B_{n!})}\geq\frac{\mu(B_{n!}\setminus B_{(n+1)!})}{\mu(B_{n!})}=\frac{2/n!-2/(n+1)!}{2/n!}=1-\frac{1}{n}\longrightarrow 1$$ and we see that the Lebesgue upper density of $E$ at $0$ is $1$.

On the other hand, if $n$ is even, then $E\cap B_{n!}\subseteq B_{(n+1)!}$. Hence $$\frac{\mu(E\cap B_{n!})}{\mu(B_{n!})}\leq\frac{\mu(B_{(n+1)!})}{\mu(B_{n!})}=\frac{2/(n+1)!}{2/n!}=\frac{1}{n+1}\longrightarrow 0$$ and we see that the Lebesgue lower density of $E$ at $0$ is $0$.

Since the upper and lower density of $E$ at $0$ differ, the density of $E$ at $0$ does not exist.

LostInMath
  • 4,528
4

For the first part, consider the region inside an angle centred in $x$ (with infinite semilines) with angle $\theta $ such that $\theta/2\pi=\alpha$. Geometrically it is obvious that the ratio is always equal to $\alpha$, and so the limit is $\alpha$.

Beni Bogosel
  • 23,381
  • 1
    For those interested in this sort of thing, an extreme example is given in the following paper: Allan deCamp, The construction of a Lebesgue measurable set with every density, Real Analysis Exchange 16 (1990-91), 344-348. – Dave L. Renfro Oct 31 '11 at 14:00
  • 1
    For a one dimensional example can I just project the circle to $\mathbb{R}$ like $[-\alpha n, \alpha n]$? – user62089 Apr 01 '13 at 16:49