I need some help with an exercise for my analysis course. Consider $Z=\{x_0,x_1,...,x_n\}$ as a Partition of the interval $[a,b]$ and upper and lower sum of a function $f: [a,b] \rightarrow ℝ$ are defined as $$O(f,Z):=\sum_{i=1}^n sup[f(\zeta)](x_i-x_{i-1})$$ $$U(f,Z):=\sum_{i=1}^n inf[f(\zeta)](x_i-x_{i-1})$$

Show that the following statements are equivalent:

a) $\forall\varepsilon>0 \exists Z: O(f,Z)-U(f,Z)<\varepsilon$

b) $\forall\varepsilon>0 \exists\delta>0\forall Z: [|Z|<\delta \implies O(f,Z)-U(f,Z)<\varepsilon]$

The direction $b) \implies a)$ is trivial in my opinion. For the other direction it seems I have not enough information to prove it.


What you need is this geometric lemma.

Lemma. Given $\epsilon>0$ and a partition $\mathbf{P}$ of $[a,b]$, there is a $\delta>0$ such that if $\mathbf{Z}$ is another partition with $|\mathbf{Z}|<\delta$, then the total length of the subintervals of $\mathbf{Z}$ that are not contained in a subinterval of $\mathbf{P}$ is less than $\epsilon$.

Proof. If $N$ is the number of points in $\mathbf{P}$, take $\delta=\epsilon/N$ $\qquad\square$.

Now, when comparing the sums with respect to $\mathbf{P}$ and $\mathbf{Z}$, there are two types of subintervals of $\mathbf{Z}$:

  1. Those containing points of $\mathbf{P}$, but whose total length is less that $\epsilon$.
  2. Those contained in subintervals of $\mathbf{P}$, what allows one to make estimates on $U(f,\mathbf{Z})-O(f,\mathbf{Z})$.

Split $U(f,\mathbf{Z})-O(f,\mathbf{Z})$ into two sums. Let $M$ be a bound of $f$. On subintervals of type 1. the sum is bounded by $2\,M\,\epsilon$. Let $I$ be a subiberval of $\mathbf{Z}$ of type 2. and let $J$ be a subinterval of $\mathbf{P}$ containing $I$. Then $$ 0\le\sup_{I}f-\inf_{I}f\le\sup_{J}f-\inf_{J}f. $$ This should be enough to allow you to finish the proof.

  • $\begingroup$ Okay, I understand the Lemma now. I've tried to split up the sums ech in the parts where I sum over all intervals which are not contained in subintervals in $P$ and over those who are contained in subintervals of $P$. I also know the the function is bounded. $\endgroup$ – hardy Apr 3 '17 at 19:38
  • $\begingroup$ I have added some more hints to finish the proof. $\endgroup$ – Julián Aguirre Apr 4 '17 at 10:23

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.