This was written in Freitag's (p196)

A continuous function $f:[a,b) \rightarrow \mathbb{C}$ is called improperly integrable iff the limit $$ \lim_{t\rightarrow b} \int_a^t f(x) \, dx $$ exists. It is (improperly) absolutely integralbe, if the above limit exists with $f$ replaced by $|f|$. Absolute integrability implies integrability (?)

What I don't understand is why the last holds bolded line holds.

My thoughts: I don't even see how this holds if $f$ were real. If we define $G(t):= \int_a^t f(x) \, dx$, then $G(t)$ is a continuous (by FTC), and bounded (by absolutely integrability) function on $[a,b)$. We need to show $\lim_{t \rightarrow b} G(t)$ exists, but it can be the case that $G(t)$ is some function like $\sin (1/x)$.


Outline: imitate the proof for absolute convergence $\implies$ convergence for sums. Write down the Cauchy criterion for $g(x) = \int_0^x \lvert f \rvert$, and use it and the triangle inequality to show that $h(x) = \int_0^x f$ is Cauchy.

  • $\begingroup$ Thanks a lot. Now I have a new question, does the Improper Integral (if exists) coincide with the Lebesgue Integral? So $ \lim_{t \rightarrow b} \int_a^t f(x) \, dx = \int_{[a,b)} f(x) \, dx . $. I think this holds in the case when $f$ is a nonnegative function as we can apply MCT or DCT, but I can't really figure it out for general case. $\endgroup$ – CL. Jun 28 '17 at 12:33
  • $\begingroup$ Yes, one can use DCT: $f \cdot 1_{[a,t)}$ is dominated by $\lvert f \rvert$ and converges to $f$ pointwise (if necessary, prove it for a sequenc $t_n \uparrow b$ and use that a function converges to a value at a point if and only if it converges on every sequence with limit that point). $\endgroup$ – Chappers Jun 28 '17 at 12:44
  • $\begingroup$ The problem with applying DCT with the given definition is we need $\int |f| \, d\mu < \infty$ ($\mu$ is Lebesgue measure.) If $f$ is improperly integrable, then we can, as you said, take an arbitrary sequence $t_n \rightarrow b$, so that $\lim_{ n \rightarrow \infty} \int_a^b f_n \, dx =: \int_a^b f \, dx $ (definition), where $f_n \rightarrow f$. BUT the problem is we cannot apply DCT or MCT. So we cannot say $\int_{[a,b) } f \, d\mu = \int_a^b f (x) \, dx $ right? $\endgroup$ – CL. Jun 28 '17 at 21:37
  • $\begingroup$ If $f$ is improperly Riemann integrable $\int_a^b f(x) \, dx $ is a Lebesgue integral if and only if $\int_a^b \lvert f(x) \rvert \, dx < \infty$ (due to how the Lebesgue integral is defined, $f$ Lebesgue-integrable requires $\lvert f \rvert$ to be as well). $\endgroup$ – Chappers Jun 29 '17 at 5:29

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.