MATH5825 Lec 1
MATH5825 Lec 1
Dmitriy Zanin
Riemann integral was defined for the functions which are continuous (or,
at least, are not very far from being continuous).
For every partition of an interval [a, b], one writes upper and lower
Darboux sums (see next slide);
Every upper Darboux sum is bigger than every lower Darboux sum
(not necessarily for the same partition);
Heuristically, value of an integral lies between that for upper and
lower Darboux sum;
As partitions get finer and finer, the difference between respective
upper and lower Darboux sums is expected to become smaller and
smaller;
One can now define Riemann integral as infimum of all upper Darboux
sums (or, equivalently, the supremum of all lower Darboux sums);
Darboux sums
Let {xk }K
k=1 be a partition of an interval [a, b]. That is,
a = x0 ≤ x1 ≤ · · · ≤ xK = b.
Definition
Upper Darboux sum for the partition {xk }K
k=1 is given by the formula
K
X −1
(xk+1 − xk ) · sup f (x).
k=0 x∈[xk ,xk+1 ]
K
X −1
(xk+1 − xk ) · inf f (x).
x∈[xk ,xk+1 ]
k=0
Theorem
If f is a continuous function on [a, b], then it is Riemann integrable.
Theorem
If f is a monotone function on [a, b], then it is Riemann integrable.
Besides that, not much can be inferred from the definition. Heuristically, a
function is Riemann integrable if it does not have too many discontinuity
points. The following theorem provides a rigorous version of this heuristic
(warning: its statement already requires some measure theory).
Theorem
A bounded function f on [a, b] is Riemann integrable if and only if the set
of its discontinuity points has measure 0.
Passing to a limit
Let
0,
x =0
fn (x) = n, 0 < x < n1
1
0, n ≤x ≤1
We have
lim fn (x) = 0 for all x ∈ [0, 1].
n→∞
Suppose you have a heap of coins (lots of different currencies) and you
want to know how much money you got. There are 2 approaches:
at every step, take a single coin, compute its value and add it to a
total value;
split the heap into groups per currency (dollars, euros, yuans, tughriks
etc.). at each step, take a group, count a number of coins there and
multiply by the exchange rate, after that add the result to a total
value;
The first process describes Riemann integral. The second process describes
Lebesgue integral.
4 if f ≥ 0 and b ≥ a, then
Z b
f (x)dx ≥ 0.
a
5 we have Z 1
1dx = 1.
0
6 if the sequence {fn }n≥1 increases and converges to f , then
Z b Z b
fn (x)dx → f (x)dx.
a a
Theorem
There exists a sequence {fn }n≥1 of Riemann integrable functions such that
fn ↑ f , where f is a Dirichlet function.
For this example, Lebesgue’s last condition fails badly — the right hand
side becomes meaningless.
Proof.
Let us write Q as a sequence {rn }n≥1 of distinct numbers. For every
n ≥ 1, define a function fn by setting
0,
x∈/Q
fn (x) = 1, x = rm , m ≤ n
0, x = rm , m > n
Each fn differs from 0 at finitely many points and is, therefore, Riemann
integrable.
We claim fn+1 ≥ fn for n ∈ N. Indeed, for x ∈ / Q, we have
fn (x) = 0 = fn+1 (x). For x ∈ Q, we have x = rm for some m ∈ N. If
n + 1 < m, then fn (x) = 0 = fn+1 (x). If n ≥ m, then fn (x) = 1 = fn+1 (x).
If n = m − 1, then fn (x) = 0 < 1 = fn+1 (x). Thus, fn (x) ≤ fn+1 (x) for all
n ≥ 1 and for all x ∈ R.
If x ∈
/ Q, then fn (x) = 0 → 0 = f (x). If x ∈ Q, then x = rm for some
m ∈ N. We have fn (x) = 1 for n ≥ m. Therefore, fn (x) → 1 = f (x) for all
x ∈ R.
Dmitriy Zanin September 12, 2024 19 / 20
What to expect from an integral?
Conclusion