Let $f$ be bounded and Riemann-integrable on $[a,b]$. Prove that $f$ is continuous at some point $x \in [a,b]$ and is also continuous on a dense subset $D$ of $[a,b].$

Let $\epsilon >0$. Since $f$ is Riemann-integrable, there exists a partition $P$ such that $U(P,f,\alpha) - L(P,f,\alpha) < \epsilon$, where $\alpha$ is an increasing function. Are we able to conclude the first part of the proof that wants us to show that $f$ is continuous at some point $x \in [a,b]$ by taking the partition $P$ to be $[a,b]$? If not, how should we proceed/conclude?

Some definitions:

A set $E$ is dense in $X$ if every point of $X$ is a limit point of $E$, or is a point of $E$ (or both).

Let $A \subset \mathbb{R}$, let $f:A \to \mathbb{R}$, and let $c \in A$. $f$ is continuous at $c$ if for any $\epsilon >0$ there exists $\delta >0$ such that $x \in A$ implies that if $|x-c| < \delta$ then $|f(x)-f(c)| < \epsilon$.

A function is Riemann-integrable if $\inf U(P,f,\alpha) = \sup L(P,f,\alpha)$, where the inf and sup are taken over all partitions.


For any real function $f$ bounded on $[a,b]$ denote

$\displaystyle U(P,f,\alpha ) = \sum_{i=1}^n M_i \Delta \alpha _i$ and

$\displaystyle L(P,f,\alpha ) = \sum_{i=1}^n m_i \Delta \alpha _i$, where

$M_i = \sup \{ f(t): x_{i-1} \leq t \leq x_i \}$ and

$m_i = \inf \{ f(t): x_{i-1} \leq t \leq x_i \}$.

A function is Riemann-integrable if $\inf U(P,f,\alpha) = \sup L(P,f,\alpha)$, where the inf and sup are taken over all partitions. (This is Rudin's definition of Riemann-integrable. $\alpha$ is an increasing function so that $\Delta \alpha _ i = \alpha (x_i) - \alpha (x_{i-1})$. For example, if $\alpha = x$, we get the definition of $\Delta x$.)

  • 1
    $\begingroup$ Sorry, what's your definition of Riemann-integrable? Where does $\alpha$ come in? $\endgroup$ – jgon Feb 3 '18 at 2:17
  • $\begingroup$ @jgon I just made an edit with clarifications. $\endgroup$ – Dominated Convergence Theorem Feb 3 '18 at 2:38
  • $\begingroup$ See math.stackexchange.com/a/519921/72031 and your second claim follows from the first. $\endgroup$ – Paramanand Singh Feb 3 '18 at 4:26
  • $\begingroup$ That sounds like the definition of Stieltjes integral, not just Riemann. $\endgroup$ – Martin Argerami Feb 4 '18 at 1:05
  • $\begingroup$ R. integrable implies continuous a.e. implies the continuity set is dense $\endgroup$ – MathematicsStudent1122 Feb 4 '18 at 3:55

Define the oscillation, $\omega_f(U)$ for an open set $U$ as $\omega_f(U)=\sup_{x\in U} f(x)-\inf_{x\in U}f(x)$. Then if $x\in \newcommand{\RR}{\mathbb{R}}\RR$, define $$\omega_f(x) = \lim_{\epsilon\to 0} \omega_f(B_\epsilon(x)).$$

It's easy to see that $f$ is continuous at $x$ if and only if $\omega_f(x)=0$. Now consider $A_n:=\omega_f\newcommand{\inv}{^{-1}}\inv([0,1/n))$ for $n > 0$. I claim that $A_n$ is open and dense for all $n$, then we'll have that the points of continuity, $\bigcap_{n\in \Bbb{N}_+} A_n$ will be dense by the Baire category theorem.

Thus we just need to show that $A_n$ is open and dense. Suppose $x\in A_n$. Then $x$ has a neighborhood $V$ such that $\omega_f(V)< 1/n$. Hence for all $y\in V$, $\omega_f(y)\le \omega_f(V)<1/n$. Thus $V$ is an open nhood of $x$ contained in $A_n$. Thus $A_n$ is open. As for density, here we need to use integrability of $f$. Suppose $(r,s)$ is a nonempty interval contained in the complement of $A_n$. Then for all $x\in (r,s)$, $\omega_f(r,s) \ge 1/n$. Now take any partition $P$ containing $r$ and $s$. Then $U(P,f,\alpha)- L(P,f,\alpha) \ge 1/n(\alpha(s)-\alpha(r))$ for any such $P$, so the Riemann integral doesn't converge. (Assuming $\alpha$ is strictly increasing. If $\alpha$ weren't strict, bad stuff could happen when it remained constant. Therefore $A_n$ is open and dense, so by the Baire category theorem the points of continuity are dense.


You can prove this by contradiction.

Suppose there is no dense subset of $[a,b]$ such that f is continuous.

There exists a point $x_1\in[a,b]$ that in a finite interval, $c>0$, and $[x_1-c,x_1+c]$ f is not continuous.

Since f is not continuous in $[x_1-c,x_1+c]$, there exist $p > 0 $ such that $|f(x)-f(t)|\geq p$ for $x,t\in[x_1-c,x_1+c]$

Then $U(P,f,\alpha)-L(P,f,\alpha) > p[\alpha(x_1+c)-\alpha(x_1-c)]$. This contradicts to f is integrable.

  • $\begingroup$ But Thomae's function is discontinuous on the rationals (which are dense), yet it is Riemann integrable. $\endgroup$ – Martin Argerami Feb 3 '18 at 20:46
  • $\begingroup$ You can see my edit. The statement is slightly different from what's in Thomae's function. $\endgroup$ – Math The Novice Feb 4 '18 at 3:49

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.