Homepage › Solution manuals › Terence Tao › An Introduction to Measure Theory › Exercise 1.1.22 (Equivalence of Riemann and Darboux integral)
Exercise 1.1.22 (Equivalence of Riemann and Darboux integral)
Let be an interval, and let be a bounded function. Show that is Riemann integrable if and only if it is Darboux integrable, and both integrals are equal.
Answers
We start with the easy one.
- Suppose that
is Darboux integrable with the Darboux integral ,
i.e., for an arbitrary
we have an upper piecewise integral with
and a lower piecewise integral with .
Since
are piecewise constant, we can find a common partition
of
which we can associate with the points
and the piecewise constant values
and
in an obvious manner. For each
choose an arbitrary tag .
We then have
In other words, we have found a tagged partition such that is -close to . Notice that if we choose any other tagged partition with the size of , then it is easy to verify (by taking common refinement again) that the inequality above still holds; thus the convergence along the filter is ensured.
- Now suppose that
is Riemann-integrable with the Riemann integral .
Fix
and choose a tagged partition
such that .
Notice that we can establish the bounds (which also happen to be upper
and lower piecewise integrals for ):
where can be chosen arbitrarily by definition. In particular, we can set such that and such that where . For both cases we have:
and
To summarize, we have found a Riemann sum which is -close to the Riemann integral , and we have constructed upper and lower piecewise integrals for which are close to the Riemann sum. Thus, by triangle inequality, we have found upper and lower piecewise integrals which are -close to . Since our choice of was arbitrary, by properties of supremum and infimum the lower Darboux integral must be equal to upper Darboux integral, both being equal to .
Comments
-
I am wonder about the backward implication. I think I have a counter example to your claim that that new partitions P' with norm less than the norm P, we retain epsilon closeness. I just considered at piece wise function f that is zero for x negative and proportional to x for x positive. Then divide up the region [-1,1] with only only one interval for [-1,0], and many intervals for [0,1]. This partition has norm unity, and we can get epsilon closeness for the integral by making the partition on the positive finer. But if we consider a uniform partition of [-1,1], say into four intervals, (so the norm of this partition is 1/2, less than the norm of the prior partition) we lose epsilon closeness. Let me know what you think, I don't want to submit it as a solution because I don't have one. I really appreciate your website.kent • 2022-04-05
Let be a box and be a bounded function. Show that is Riemann integrable if and only if it is Darboux integrable, in which case the Riemann integral and the Darboux integral are equal.
Fix an arbitrary . Let be Riemann integrable on . Then there exists a such that
for all tagged partitions with .
Let be a tagged partition with , such that for every we have .
Then define as for . Now for all we have and thus pointwise and is piecewise constant.
Similarly, let be a tagged partition with , such that for every we have . Define as for . Now for all we have and thus pointwise and is piecewise constant.
Then
so that is Darboux integrable.
Given a , denote as the set of piecewise continuous functions on with respect to an arbitrary partition such that and with the property for that
for any .
Let .
So we have and that gives the lowest possible Darboux integral over for any piecewise continuous function with respect to a partition with , but that still has the property that it can serve as a tagged partition for , since for some (or at least is -close to such an ).
We have to prove that
Let be piecewise constant on and be such that . This is guaranteed by being Darboux integrable. Let be the partition alongside which is piecewise constant with . Then choose like above. It is possible that by the difference in the partitions chosen. Now consider the refinement partition with and consider .
Assume to the contrary that . Then for some for which for , we must have , because otherwise by monotonicity we would get
But we also know for some and then , which is a contradiction to . Therefore,
.
The proof for goes analogously, with similar notation.
Let be an arbitrary tagged partition with . Then for any box we must have for some , that is
for all , from which we can deduce
that is
in other words
for all tagged partitions with .