2

The question is most probably simple, and is related to the theorem 1.17 in Tom Apostol's calculus.

Can we prove subinterval integrability theorem: "Let $a < c < b$ be real numbers. If $\int_{a}^{b} f(x) dx$ exists, then $\int_{a}^{c} f(x) dx$ also exists"?

It is easy to prove if we know that two integrals on exist, as the theorem 1.17 implies, but I wonder if the theorem above also holds (e.g. if the integrability on a larger interval implies integrability on a subinterval). If I try to prove that the way 1.17 was proved (page 86), as follows:

$$ \int_{a}^{b} f(x) dx \iff \sup\left\{ \int_{a}^{b}s \biggm{\vert} s \leq f\right\} = \inf\left\{ \int_{a}^{b}t \biggm{\vert} f \leq t\right\} $$

Where $s$ and $t$ are step functions (more details can be found in theorem 1.9 in the Calculus book, section 1.17). By the additive property of supremum and infimum (T I.33), we have

$$\begin{align} \sup\left\{ \int_{a}^{b}s \biggm{\vert} s \leq f\right\} &= \sup\left\{ \int_{a}^{c}s \biggm{\vert} s \leq f\right\} + \sup\left\{ \int_{c}^{b}s \biggm{\vert} s \leq f\right\}\\ &= \inf\left\{ \int_{a}^{c}t \biggm{\vert} f \leq t\right\} + \inf\left\{ \int_{c}^{b}t \biggm{\vert} f \leq t\right\}\\ &= \inf\left\{ \int_{a}^{b}t \biggm{\vert} f \leq t\right\} \end{align}$$

I would be done if it would somehow follow from the above that

$$ \sup\left\{ \int_{a}^{c}s \biggm{\vert} s \leq f\right\} = \inf\left\{ \int_{a}^{c}t \biggm{\vert} f \leq t\right\} $$

But it does not, so I'm not sure how to proceed, and if it is possible to be proved, without using more advance techniques than what he covered until that point in the book.

A previous answer here mentions that the theorem holds, but I'm not sure what method would the commenter use to do that.

Thanks!

shoteyes
  • 1,435
S11n
  • 978
  • 1
    I don’t know if Apostol ever proves that fact across both volumes of the book considering the proofs of later theorems often state, “if $f$ is integrable on $[a, x]$ for every $x\leq b.$” The accepted answer in the linked question relies on the fact that the product of integrable functions is integrable, which may be tough to prove only using what is mentioned in the book up to that point. That’s probably the reason for the assumption that at least two of the integrals $\int_a^b f,$ $\int_a^c f,$ and $\int_c^b f$ exist in the statement of Theorem 1.17. – shoteyes Jul 03 '21 at 14:56
  • Thanks for the comment! Yes, probably that's what I'm missing at this point. If that would be true (that the product of 2 integrable functions is integrable) and I understand correctly, then the $inf$ and $sup$ above between $(c, b)$ would be 0, so the last equality would hold and it would be easy to prove. – S11n Jul 05 '21 at 06:51

0 Answers0