measure theory: simple example, testing understanding of egoroff's theorem statement

48 Views Asked by At

I'm being introduced to measure theory, in particular, egoroff's theorem. The version stated in my book considers extended real valued functions and reads as follows:

Let $\Omega \subset \mathbb{R}^n$ be $\mu$ measurable with $\mu(\Omega)<\infty$. Consider a sequence of $\mu$ measurable functions $f_k: \Omega \to \overline{\mathbb{R}}$ converging pointwise $\mu$ almost everywhere to a $\mu$ measurable function $f: \Omega \to \overline{\mathbb{R}}$ with $|f(x)|<\infty$ $\mu$ almost everywhere. Then $$\forall \delta >0 \exists A \subset \Omega \ \mu \ meas.: \mu(A\backslash \Omega) < \delta \ and \ f_k \to f \ unif. \ on \ A$$ Note that if $\mu$ is Radon, then we can choose $A$ to be compact.

The example that came to my mind is probably the first one everyone encounters as a sequence of functions where the convergence is pointwise but not uniform. Take $f_n: x \in [0;1] \mapsto x^n \in \mathbb{R}$ converging pointwise to the zero function on $[0;1)$ and to $1$ for $x=1$. My intuition of the convergence not being uniform, is that the point $x=1$ is the "issue", meaning the convergence is uniform on $[0;1)$. This would be coherent with egoroff's theorem, where $A=\{1\}$ and thus (take $\mathcal{L}^1$ as your measure) $\mathcal{L}^1([0;1]\backslash A)=0< \delta \ \ \forall \delta >0$ and therefore $x^n$ converging uniformly to the zero function on $A=[0;1)$.

My questions are

  1. Why is my intuition wrong? Thinking about it, $f_k|_{[0;1)}$ still does not converge uniformly to the zero function, or am I mistaken?
  2. If that is the case, how is egoroff's theorem applicable in this case? What is my A so that the convergence is uniform on A?

Thank you for the help!

1

There are 1 best solutions below

1
On
  1. No, $f_n$ does not converge uniformly on $[0,1).$
  2. Your $A$ could be $[0,1-\delta ]$. Then, $f_n$ converges uniformly on $A$ and $\mu([0,1]\setminus A) = \delta.$

Egoroff's theorem says you can take away a set of arbitrarily small positive measure and recover uniform convergence. So you shouldn't necessarily expect to be able to recover it by only deleting the measure-zero set $\{1\}$ from the "problem area".