Here's an interesting problem I stumbled upon while I was pondering on the $(\varepsilon,\delta)$ definition of a limit:
Suppose that $\lim_{x\to a}f(x)=L$ ($f$ is a real function). Fix some $\varepsilon>0$ and let $S_\varepsilon$ be the set $$\{\delta\in\mathbb{R}:|f(x)-L|<\varepsilon\text{ is true for every $x\in\text{dom}[f]$ satisfying }0<|x-a|<\delta\}$$ Does $S_\varepsilon$ always contain an interval? If so, can we do better and conclude that $S_\varepsilon$ is an interval that is either $(0,b)$, $(0,b]$, or $(0,\infty)$ for some $b>0$?
I’m pretty sure the answer to the first question is yes. Here's my argument:
If $\lim_{x\to a}f(x)=L$, then $S_\varepsilon$ is not empty because the $(\varepsilon,\delta)$ definition of a limit implies that there is a $\delta>0$ such that
$$|f(x)-L|<\varepsilon\text{ is true for every }x\in\text{dom}[f]\text{ satisfying } 0<|x-a|<\delta$$
If we then fix any $\delta_1\in S_\varepsilon$, we can infer that every $s\in(0,\delta_1)$ will be an element of $S_\varepsilon$, since
\begin{align} x\in\text{dom}[f]\text{ and }0<|x-a|<s\text{ and }s<\delta_1 &\implies x\in\text{dom}[f]\text{ and }0<|x-a|<\delta_1\\ &\implies|f(x)-L|<\varepsilon \end{align} It follows $(0,\delta_1)\subseteq S_\varepsilon$, so $S_\varepsilon$ contains an interval.
Does this argument work? If so, is the stronger claim mentioned in the problem statement also demonstrably true? After spending some time visualizing the $(\varepsilon,\delta)$ definition for constant functions and functions which are bounded/unbounded at the endpoints of their domain, I suspect that the answer is yes. Unfortunately, I have no idea how to prove this. Any help is appreciated.
Looks like you have the gist of it, but you might be bogged down by some unnecessarily complicated notation. How about this:
Let $\delta > 0$ be such that for any $x$ that satisfies the equation \begin{equation} \tag{1} \left| x - a \right| < \delta, \end{equation} we have \begin{equation} \tag{2} \left|f(x) -L \right| < \varepsilon \end{equation}
Let $0<\delta' < \delta$, if $\left| x- a \right| < \delta'$, then, since $\delta' < \delta$, we have $\left| x -a \right| < \delta' < \delta$ i.e. $x$ satisfies equation (1). This implies that equation (2) also holds.
In particular, this implies that if $\delta > 0$ is in $S_\varepsilon$, then $\left(0, \delta \right) \subset S_\varepsilon$. EDIT: Call this Fact 1.
EDIT: $S_\varepsilon$ is an interval:
Ether $S_\varepsilon$ has a supremum or not. If the supremum $s$ exists, then by Fact 1 we have that $S_\varepsilon$ is either $(0, s)$ or $(0, s]$.
If $S_\varepsilon$ does not have a supremum, for any $M > 0$ there is an element $\delta > M$ in $S_\varepsilon$. Hence, by Fact 1, the set $S_\varepsilon$ contains the interval $(0, M)$ for arbitrarly large $M$'s. Thus $S_\varepsilon = (0,\infty)$.