Given a Banach space $(X,||\cdot||_{X})$ over $\mathbb{C}$ and a Riemann Integrable $X$-valued function $f$ on $[a,b]$, I am trying to show that $||f||_{X}$ is a real-valued, Riemann Integrable function on $[a,b]$.

Since $||f||_{X}$ is real-valued, we may show that it's R.I. by showing that for every $\epsilon>0$, there exists a partition of $[a,b]$, $P=(p_{k})_{k=0}^{n}$, such that: \begin{equation} \sum_{k=1}^{n}\Big{(}\underset{p_{k-1}\leq x\leq p_{k}}\sup||f(x)||_{X}-\underset{p_{k-1}\leq x\leq p_{k}}\inf||f(x)||_{X}\Big{)}\triangle P_{k}<\epsilon \end{equation} In particular, fixing $\epsilon>0$, we may find $\delta>0$ such that for any two tagged partitions of $[a,b]$, $(P,Q),(P',Q')$ with $||P||,||P'||<\delta$, the norm of the difference of the Riemann Sums on these tagged partitions is less than $\epsilon$. (This is the "Cauchy Criterion" as $f$ is R.I.)

Now, choosing $P=(p_{k})_{k=0}^{n}$ with $||P||<\delta$, we may approximate the above sum as: \begin{multline*} \sum_{k=1}^{n}\Big{(}\underset{p_{k-1}\leq x\leq p_{k}}\sup||f(x)||_{X}-\underset{p_{k-1}\leq x\leq p_{k}}\inf||f(x)||_{X}\Big{)}\triangle P_{k} \\ \leq \sum_{k=1}^{n}\Big{(}||f(t^{\sup}_{k})||_{X}-||f(t^{\inf}_{k})||_{X}\Big{)}\triangle P_{k} + 2\epsilon(b-a)\end{multline*} via the approximation property of the sup and inf on each interval $[p_{k-1},p_{k}]$. This estimate may then be bounded by: $$\sum_{k=1}^{n}||f(t^{\sup}_{k})-f(t^{\inf}_{k})||_{X}\triangle P_{k} +2\epsilon(b-a)$$ This is as far as I've gotten. There is a similar question on this site where they try to bound the last sum above, but I am not sure their answer is correct. Could someone give me a hint on where to go from here?


I would say that since $f$ is Riemann-integrable, it has at most a countable number of discontinuities (according to Froda's theorem). Since the norm is continuous, that would make $||f||_{X}$ continuous except maybe for a countable number of points, which implies that $f$ is Riemann-integrable (since it is in particular continuous almost everywhere).

  • $\begingroup$ Froda's Theorem does not work here. It describes the set of discontinuities of monotone real-valued functions. Above, $f$ is a Banach-valued function where the Banach space is an arbitrary Banach space over $\mathbb{C}$, so we may not even have the notion of monotonicity of $f$. More generally, one could investigate whether or not Lebesgue's criterion for Riemann-Integrability of real-valued functions will hold in arbitrary Banach spaces, and the answer is no. $\endgroup$ – JWP_HTX Oct 9 '17 at 22:56
  • $\begingroup$ I believe that Froda's Theorem extends to functions $[a,b] \rightarrow B$ where $B$ is a Banach space and $f$ is a uniform limit of step functions ($f$ thus admits a left- and right-limit at each point, and the same proof as for monotone functions applies). This is the case here since $f$ is Riemann-integrable with values in a Banach space (and thus can be approached uniformly by step functions). Please do correct me if I am wrong, thanks! $\endgroup$ – Julien Oct 10 '17 at 16:11
  • $\begingroup$ Hi Julien, looking at the proof of Froda's Theorem on Wikipedia, it relies on the notion that we can compare values of $f$, which we don't have in an arbitrary Banach space. I am skeptical that such a proof could extend to Banach spaces, especially since Riemann-Integrability in an arbitrary Banach spaces does not imply that the set of discontinuities has measure zero. Could you maybe provide a link to a proof of Froda's Thm for Banach spaces? $\endgroup$ – JWP_HTX Oct 10 '17 at 17:44
  • $\begingroup$ If $f: [a;b] \rightarrow M$ ((M,d) any metric space) admits a jump discontinuity of size $\epsilon>0$ at point $x_{0}$, then $\exists \eta>0, \forall x, d(x_{0},x) \leq \eta \Rightarrow Variation(f(]x_{0}-\eta;x_{0}[) < \epsilon$ and $Variation(f, ]x_{0};x_{0}+\eta[) < \epsilon$. This shows that $x_{0}$ is the only jump discontinuity with size $>\epsilon$ on $]x_{0}-\eta;x_{0}+\eta[$, thus the set of jump discontinuities of $f$ is at most countable. Now every Riemann-integrable function in a Banach is a uniform limit of step functions (and thus has a left/right limit at every point)... $\endgroup$ – Julien Oct 10 '17 at 20:46
  • $\begingroup$ "Riemann-Integrability in an arbitrary Banach spaces does not imply that the set of discontinuities has measure zero": in this case maybe we do not use the same construct of the Riemann integral in a Banach space? I see it as the sum of step functions, extended to the closure of the set of all step functions with values in the Banach space (for the norm $||.||_{\infty}$). $\endgroup$ – Julien Oct 10 '17 at 20:56

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.