Proof Question: Continuous Functions on Compact Domains are uniformly continuous

compactnessproof-explanationreal-analysisuniform-continuity

So, I'm going over the proof of the following theorem in Real Analysis, but I am unable to comprehend a step in the proof. The theorem is as follows:

$\textbf{Theorem:}$ Let $f$ be a continuous function on a compact domain $K \subset \mathbb{R}$. Then $f$ is uniformly continuous on $K$

Proof: Pick $\epsilon > 0$. By the definition of continuity there is for each $x \in K$ a number $\delta_x > 0$ such that if $|x-t| < \delta_x$ then $|f(t) – f(x)| < \frac{\epsilon}{2}$. Observe that the intervals $I_x = (x – \frac{\delta_x}{2}, x + \frac{\delta_x}{2})$ form an open covering of $K$. Then, since $K$ is compact, we may extract a finite subcovering
$$I_{x_1}, \ldots, I_{x_m}$$

Everything is good so far, but then it says:

Let $\delta = \min\{\frac{\delta_{x_1}}{2}, \ldots, \frac{\delta_{x_m}}{2}\}$

This is where I get confused. I know that the $\delta_{x_i}$ are presumably positive numbers, but where do these numbers live? How are we getting them? Are we assuming that each of the sets $I_{x_i}$ are simply equal to the intervals $(x_i – \frac{\delta_{x_i}}{2})$? If so, why can we assume this about the structure of the $I_{x_i}$?

Best Answer

The second sentence shows you the property that $\delta_x$ must satisfy, given $x \in K$. For each $x \in K$, you now have an associated $\delta_x > 0$ such that $$y \in (x - \delta_x, x + \delta_x) \implies |f(y) - f(x)| < \frac{\varepsilon}{2}.$$ The proof goes on to call the interval $(x - \delta_x, x + \delta_x)$ by the name $I_x$. So, for each $x \in K$, you also have an interval $I_x$, containing $x$. So, using $K$ as an index set, you have a family of open sets $(I_x)_{x \in K}$ with the union containing $K$. We need this to use compactness.

Compactness implies the existence of a finite subcover of this open cover. There must be a finite subset of points from $K$ (i.e. a set of the form $J = \{x_1, \ldots, x_m\} \subseteq K$ for some $m$) such that $$K \subseteq \bigcup_{x \in J} I_x = \bigcup_{n = 1}^m I_{x_n}.$$

But remember, we have $\delta_x$ associated to any given $x \in K$, including the elements of $J$! As such, we have a $\delta_{x_n}$ associated to $x_n$ for all $n = 1, \ldots, m$.

You could also think of $\delta_{x_n}$ as half the diameter of $I_{x_n}$, if that helps.