Riemann Integration - Random Walks
Riemann Integration - Random Walks
Contents
Riemann integral
Deficiencies of the Riemann integral
The first part of the book introduces the Riemann integral. The Riemann integral is good
enough for many use cases, but it also has important limitations. We first review the definition
of the Riemann integral and its properties. Then we discuss the limitations of th Riemann
integral and show important ways in which it falls short. This will lead to the idea of the outer
measure as a more appropriate definition of length, however we will see that this too fails to
capture important properties that we would like it to have. This leads on to the Lebesgue
measure, the Lebesgue integral and more generally Lebesgue integration.
Riemann integral
We begin with a few definitions needed to write down the definition of the Riemann integral.
Given a function f : [a, b] → R, we will define its Riemann integral as the limit of an
approximation of the area under the curve of f by rectangles which partition [a, b]. A partition
of [a, b] is defined as follows.
Definition 50 (Partition)
With these definitions in place we can define upper and lower Riemann sums. A lower (upper)
Riemann sum is an approximation of the area under the curve of f by rectangles below (above)
the curve.
n
L(f, P , [a, b]) = ∑(x i − x i−1 ) inf f,
x∈[x i−1 ,x i ]
i=1
n
U(f, P , [a, b]) = ∑(x i − x i−1 ) sup f.
i=1 x∈[x i−1 ,x i ]
A very useful intermediate result is that refining a partition of [a, b], i.e. adding more points to
the partition, can only increase the lower Riemann sum and reduce the upper Riemann sum.
L(f, P , [a, b]) ≤ L(f, P ′ , [a, b]) ≤ U(f, P ′ , [a, b]) ≤ U(f, P , [a, b]).
Another useful inequality which is derived directly from the previous one, is that any lower
Riemann sum of a function is less than or equal to any of its upper Riemann sums.
With these definitions and results in place, we are ready to define lower and upper Riemann
integrals. The lower and upper Riemann integral can be thought of as a bound to the best
approximation to the area under the curve of f by rectangles below and above the curve,
respectively.
where the supremum and infimum are taken over all partitions P of [a, b].
From the definition of the lower and upper Riemann integral, and our previous results, it follows
that the lower Riemann integral is less than or equal to the upper Riemann integral.
With these definitions and results in place, we are ready to define the Riemann integral.
by
b
∫ f = L(f, [a, b]) = U(f, [a, b]).
a
The last result in this section concerns a frequently used estimate for the Riemann integral of a
function.
b
(b − a) inf f ≤ ∫ f ≤ (b − a) sup f.
x∈[a,b] a x∈[a,b]
Riemann integration does not handle functions with many discontinuities. While functions
with a finite number of discontinuities are Riemann integrable, functions with an infinite
number of discontinuities need not be.
Riemann integration does not handle bounded functions. While the Riemann integral can
be extended to handle some unbounded functions by taking limits of the integral
endpoints, this does not work well for certain functions.
Riemann integration does not work well with limits. In particular, the limit of a sequence of
Riemann integrable functions need not be Riemann integrable, and even if it is, the integral
of the limit need not be the limitSkip to main
of the content
integrals.
The next four examples illustrate these issues. The first is an example of a bounded function
which is not Riemann integrable.
1 if x ∈ Q,
f(x) = {
0 if x ∈ R ∖ Q,
n n
U(f, P , [0, 1]) = ∑(x i − x i−1 ) sup f = ∑(x i − x i−1 ) = 1.
i=1 x∈[x i−1 ,x i ] i=1
n n
L(f, P , [0, 1]) = ∑(x i − x i−1 ) inf f = ∑ 0 = 0.
x∈[x i−1 ,x i ]
Print to PDF i=1 i=1
Since this holds for any partition P of [0, 1], we have U(f, [0, 1]) = 1 and
L(f, [0, 1]) = 0. Therefore, f is not Riemann integrable.
As we saw, the function f in the previous example is bounded, but not Riemann integrable.
Further, the Riemann integral does not work for functions which are unbounded, as shown by
the next example.
1
√x
if x ∈ (0, 1],
f(x) = {
0 if x = 0,
would give U(f, [0, 1]) = ∞ for every partition P of [0, 1]. A standard way to handle
unbounded functions with the Riemann integral is to take limits of the integral
endpoints. For example, for this particular function, we could define
1 def 1
∫ f = lim ∫ f,
0 a↓0 a
where the limit a ↓ 0 means that a approaches 0 from above. Then, we would have
1 1
∫ f = lim ∫ f = lim (2 − 2√a) = 2.
0 a↓0 a a↓0
However, as the next example shows, the above approach does not always work.
Let r 1 , r 2 , … be a sequence which includes all the rational numbers in (0, 1) exactly
once and that includes no other numbers. Define the functions f n : [0, 1] → R as
1
√x−r n
if x ∈ (r k , 1],
f n (x) = {
0 if x ∈ [0, r n ],
∞
f n (x)
f(x) = ∑ .
n=1 2n
While each of the functions above has area less than 2, looking at the expression that
defines f, we would expect the area under the curve of f to be less than 2. However,
f is not Riemann integrable, and further, it is not handled by the approach we outlined
in the previous example. That is because every nonempty open subinterval of [0, 1]
contains at least one rational number, so f is unbounded on every such subinterval.
Lastly, Riemann integration does not work well with pointwise limits of functions, as shown by
the next example.
Let r 1 , r 2 , … be a sequence which includes all the rational numbers in (0, 1) exactly
once and that includes no other numbers. Define the functions f n : [0, 1] → R as
1 if x ∈ {r 1 , … , r n },
f n (x) = {
0 otherwise,
1 if x ∈ Q,
f(x) = {
0 otherwise.
Each f n is Riemann integrable and f n (x) → f(x) for all x ∈ [0, 1], but f is not
Riemann integrable – this was shown in the first example in this section.
In general, because analysis is concerned with results about limits, we would like to be able to
exchange the order of limits and integrals. While, as shown above, we cannot do this with the
Riemann integral in general, it is possible if the limit function exists and is Riemann integrable.
|f n (x)| ≤ M,
for all k ∈ N and x ∈ [a, b]. Suppose lim n→∞ f n (x) exists for each x ∈ [a, b].
Define f : [a, b] → R as
b b
∫ f = lim ∫ fn .
n→∞
a a
We do not provide a proof for this result here. However, we note that this result suffers from the
issue that the limit function f is required to be Riemann integrable. Ideally, given a good
definition of integrals, we would like this to follow directly from the other assumptions in this
theorem. The measure theoretic definition of the integral, which is developed in the next
sections, yields an exchange theorem which does not suffer from this issue.