Suppose $(X,d)$ is a metric space and a real-valued function $f$. Denote the oscillation of $f$ restricted to a closed subset $F \subseteq X$ at a point $x \in F$ by $$oscf|_{F}(x)=\inf_{U \ni x \space open} \sup \{ d(f(x_1), f(x_2)):x_1,x_2\in U \cap F \}$$

Claim: $oscf|_{F}$ is upper semicontinuous.

Definition: $oscf|_{F}:X\rightarrow \mathbb{R}$ is upper semicontinuous if for all real number $c$, the set $\{ x \in X: oscf|_{F}(x) < c \}$ is open in $X$.

My attempt: Let a real number $c$ be given and $oscf|_{F}(y) <c.$ By definition of infimum, there exists an open neighourhood $U$ of $y$ such that $\sup \{ d(f(x_1), f(x_2):x_1,x_2 \in U\cap F \} <c$. If $z \in U\cap F$, then $U$ is an open neighbourhood of $z$. Therefore $oscf|_{F}(z) <c.$ Hence, $U \cap F \subseteq \{ x \in F: oscf|_{F}(x) < c\}$ and $U \cap F$ is open in $F$. So the set $\{ x \in F: oscf|_{F}(x) < c\}$ is open in $F$ and $oscf|_{F}$ is upper semicontinuous.

Is my proof correct?


The structure of your proof is correct. You just need to be more precise about a few details and in showing that $osc\,f|_F(z) < c.$

Suppose $y \in \{x \in X: osc \, f|_F(x) < c \},$ with $osc \, f|_F(y) = a < c.$ Choosing $b$ such that $a < b < c$, there exists an open neighborhood $U$ of $y$ such that $\sup\{d(f(x_1),f(x_2)): x_1, \, x_2 \in U \cap F \} \leqslant b$.

Consider any $z \in U \cap F$. Since $U$ is open there exists an open neighborhood $V$ of $z$ such that $V \subset U$.

If $z_1, z_2 \in V \cap F \subset U \cap F$, then $d(f(z_1),f(z_2)) \leqslant \sup\{d(f(x_1),f(x_2)): x_1, \, x_2 \in U \cap F \} \leqslant b$ and, hence,

$$\sup \{d(f(z_1),f(z_2)): z_1,\, z_2 \in V\cap F \} \leqslant b, $$

which implies

$$osc \, f|_F(z) = \inf_{V, \, z \in V} \,\sup\{d(f(z_1),f(z_2)): z_1,\, z_2 \in V\cap F \}\leqslant b < c.$$

Thus, $U \subset \{x \in X: osc \, f|_F(x) < c \} $ and $\{x \in X: osc \, f|_F(x) < c \} $ is open in $X$.

| cite | improve this answer | |
  • $\begingroup$ I think all your $d(x_1,x_2)$ should be $d(f(x_1),f(x_2))$ instead. $\endgroup$ – Idonknow Apr 29 '17 at 2:33
  • $\begingroup$ Yes -- bad typo. Will fix. $\endgroup$ – RRL Apr 29 '17 at 2:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.