Definition of a lim inf of a function over a metric space.

metric-spacesreal-analysis

I'm reading a book, and it defines the following definition:

Let $f:X \to [-\infty,+\infty]$ be a function , where $(X,d)$ is some metric space.

Define $\lim\inf\limits_{y\to x}f(y) :=\inf \{\lim\inf_n (f(x_n)):x_n\to x\}$.

Note that we can take $x_n =x$ for all $n\in \Bbb N$.

From the definition we have that for all $\epsilon \gt0$ there is some $x'\in X$ s.t $lim\inf_{y\to x}f(y) -\epsilon \lt f(x')$.

Now, the author states that

$\lim\inf\limits_{y\to x}f(y) = \sup\limits_{U \in N(x)} \inf\limits_{y \in U} f(y)$

and that in tern imply that $$lim \ inf_{y\to x}f(y) = \sup\limits_{\delta\gt 0}\ \inf\limits_{y:d(x,y) \lt \delta} f(y)$$

I couldn't understand why this equality holds. I was only able to show that $RHS\le LHS$.

I would really appreciate help proving this equality.

Thanks

Best Answer

For each $n=1,2,...$, choose $y_{n}$ such that $d(x,y_{n})<1/n$ and $\inf_{y:d(x,y)<1/n}f(y)\leq f(y_{n})\leq\inf_{y:d(x,y)<1/n}f(y)+1/n$. Note that $(\inf_{y:d(x,y)<1/n}f(y))_{n=1}^{\infty}$ is monotone, then $\lim_{n}f(y_{n})=\lim_{n}\inf_{y:d(x,y)<1/n}f(y)$. Note that we also have $f(y_{n})\leq\inf_{y:d(x,y)<1/n}f(y)+1/n\leq\sup_{\delta>0}\inf_{y:d(x,y)<\delta}f(y)+1/n$ and hence $\sup_{\delta>0}\inf_{y:d(x,y)<\delta}f(y)\geq\lim_{n}f(y_{n})=\liminf_{n}f(y_{n})\geq\inf\{\liminf_{n}f(x_{n}): x_{n}\rightarrow x\}$.

On the other hand, for any $x_{n}\rightarrow x$ and $\delta>0$, then some $n_{0}$ is such that $d(x_{n},x)<\delta$ for $n\geq n_{0}$, then $f(x_{n})\geq\inf_{y:d(x,y)<\delta}f(y)$ for all such $n$ and hence $\liminf_{n}f(x_{n})\geq\inf_{y:d(x,y)<\delta}f(y)$. Since this is true for all $\delta>0$, we have $\liminf_{n}f(x_{n})\geq\sup_{\delta>0}\inf_{y:d(x,y)<\delta}f(y)$, and finally $\inf\{\liminf_{n}f(x_{n}): x_{n}\rightarrow x\}\geq\sup_{\delta>0}\inf_{y:d(x,y)<\delta}f(y).$