If $f$ is a nondecreasing function defined on $\mathbb{R}$, then the “length” of a half-open interval $(a, b]$, denoted by $\alpha_f((a;b])$ can be defined by $\alpha_f((a;b]) = f(b) - f(a)$

The Lebesgue-Stieltjes outer measure of an arbitrary set $E \subset \mathbb{R}$ is defined by $\lambda(E) = inf \displaystyle\left\{\sum_{h_k \in F}\alpha(h_k) | h_k = (a_k; b_k], E \subset\bigcup_{h_{k} \in F} h_k\right\}$* $F$ is a countable collection of half-open intervals $h_k$ of the form $(a_k; b_k]$. $\lambda$ is the Lebesgue- Stieltjes measure generated by $f$.(Modern real analysis, William P. Ziemer, pg. 96)

Question: Let $f : \mathbb{R}\rightarrow \mathbb{R}$ be a nondecreasing function and let $\lambda_f$ be the Lebesgue- Stieltjes measure generated by $f$. Prove that $\lambda_f(\{x_0\}) = 0$ if and only if f is left continuous at $x_0$.

My demonstration $(\Rightarrow)$ Let $\epsilon > 0$ there exist a countable collection $F$ of half-open intervals $h_k$ of the form $(a_k; b_k]$ such that $0 < \displaystyle\sum_{h_k \in F}\alpha_f(h_k) < \epsilon$ and $\displaystyle{x_0} \in \bigcup_{h_k \in F} h_k$. Then ${x_0} \in h_{k_i}$ for some $h_{k_i} \in F$, so $x_0 \in (a_{k_0}; b_{k_0}]$; take $\delta= |a_{k_0} - x_0|$, for any $x \in (x_0 - \delta; x_0]$ we have $f(x_0) - f(x)\leq \alpha_f((x_0 - \delta; x_0]) \leq \alpha_f(h_{k_i}) \leq \displaystyle\sum_{h_k \in F}\alpha_f(h_k) < \epsilon$ because $f$ is nondecreasing. Therefore $f$ is left continuous at $x_0$.

($\Leftarrow$) Let $\epsilon > 0$ there exist $\delta > 0$ such that for $x \in (x_0 - \delta; x_0]$ then $\alpha_f((x, x_0]) = f(x_0) - f(x) < \epsilon$. As ${(x, x_0]}$ is a enumerable collection of half-open intervals such that $\{x_0\} \subset (x ; x_0]$ then by the definition * $\lambda(\{x_0\}) = 0$ $\blacksquare$

Any mistake?


I think you can do it faster: the sequence $$ {\left\{\left(x_0 - \frac{1}{n} , x_0\right]\right\}}_{n = 1}^{\infty} $$ is contractive, so, using properties about measures, $$ {\lambda}_f(\{x_0\}) = {\lambda}_f\left(\bigcap_{n = 1}^{\infty} \left(x_0 - \frac{1}{n} , x_0\right]\right) = \lim_{n \to \infty} {\lambda}_f\left(\left(x_0 - \frac{1}{n} , x_0\right]\right) = $$ $$ = \lim_{n \to \infty} \left(f(x_0) - f\left(x_0 - \frac{1}{n}\right)\right) = f(x_0) - f(x_0^-)\mbox{.} $$ Now, it's trivial: if ${\lambda}_f(\{x_0\}) = 0$, then $f(x_0) = f(x_0^-)$, so $f$ is continuous from the left at $x_0$, but $f$ is also continuous from the right at $x_0$ because it's a distribution function (and $f$ is continuous from the right on $\mathbb{R}$ by definition of distribution function); definitely, $f$ is continous at $x_0$ (because it's continuous from the left and from the right at $x_0$). If $f$ is continuous at $x_0$, in particular it's continous from the left at $x_0$, so, by the equality ${\lambda}_f(\{x_0\}) = f(x_0) - f(x_0^-)$, it's clair that ${\lambda}_f(\{x_0\}) = 0$.


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.