Lecture Notes on Measure Theory and Integration
1 — Introduction
Joel Feldman
University of British Columbia
September 4, 2019
1
1 Introduction
1.1 What are “Measure Theory and the Lebesgue Integral”?
Rb
In introductory Calculus courses a f (x) dx is invariably defined to be the Riemann
integral. The Lebesgue integral is an alternative definition. For simplicity, assume
that a < b and f (x) ≥ 0.
Rb
• Under the Riemann integral definition of a f (x) dx,
◦ for each n ∈ N, we subdivide the domain of integration into n (for example,
equal) subintervals and correspondingly subdivide
R = (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x)
into n vertical strips (x, y) xi−1 ≤ x ≤ xi , 0 ≤ y ≤ f (x) , 1 ≤ i ≤ n.
y
y = f (x)
x0 = a x1 x2 x3 x4 x5 x6 x7 x8 = b x
◦ Then for each 1 ≤ i ≤ n, we pick some x∗i in the ith subinterval and approximate
the area of the ith strip by the area of the rectangle which has the same width
as the strip and which has height f (x∗i ).
◦ In this way we approximate the area of R by ni=1 f (x∗i )(xi − xi−1 ).
P
Rb
◦ By definition, the Riemann integral a f (x) dx is the limit of this expression as
n → ∞.
Introduction 2 September 4, 2019
Rb
• Under the Lebesgue integral definition of a f (x) dx,
◦ for each n ∈ N, we subdivide the y-axis, rather than the x-axis, into subintervals
of length n1 and correspondingly subdivide
R = (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x)
into pieces Ri = (x, y) ∈ R ni ≤ f (x) ≤ i+1
n
y
y = f (x)
(i+1)/n
i/n
Xi Xi Xi Xi x
◦ Denote by
Xi = f −1 [i/n, (i+1)/n) ∩ [a, b]
the set of x’s in [a, b] were f takes values in the interval [i/n, (i+1)/n) . On Xi
we approximate f by ni . So we approximate the area of Ri by the height ni ,
times its width, which is the “size” of Xi , given by the sum of the lengths of
the various components of Xi . We use m(Xi ) to denote that size, which we call
the “measure” of Xi .
◦ In this way we approximate the area of R by i ni m(Xi ).
P
Rb
◦ By definition, the Lebesgue integral a f (x) dx is the limit of this expression
as n → ∞.
◦ The main step in filling out this outline of the definition of the Lebesgue integral
is to make precise what we mean by “the size of a set”. That’s “measure
theory”.
Introduction 3 September 4, 2019
1.2 Why Bother with Measure Theory?
• We will eventually1 see the following theorem, which gives us some info regarding
the relationship between the Riemann and Lebesgue integrals.
Theorem. Let a < b be real numbers and let f : [a, b] → R be bounded.
(a) If f is Riemann integrable, then f is Lebesgue measurable and
Z a Z
R f (x) dx = f (x) dm(x)
b [a,b]
(b) f is Riemann integrable if and only if x ∈ [a, b] f is not continuous at x
has Lebesgue measure zero.
In part (a), the integral on the left is the Riemann integral and the integral on the
right is the Lebesgue integral. We have not yet defined “Lebesgue measure zero”.
Any set which contains at most countable many points has Lebesgue measure zero.
• The measure theory approach to integration is much more technically demanding
than the Riemann integral approach, and the Riemann integral works perfectly
well with even slightly smooth integrands. So if you work only with, for exam-
ple, piecewise continuous functions — and that is the case for much of “classical”
applied mathematics and “classical” physics (for example classical mechanics, elec-
tricity and magnetism, fluid dynamics) — there is no point is learning measure
theory.
• But if you have to deal with very irregular functions — and that is the case in
quantum mechanics and option pricing using stochastic differential equations —
the Lebesgue integral often works well even when the Riemann integral crashes
and burns.
• For example view
n Z 1 o
2
LR ([0, 1]) = f : [0, 1] → C f is Riemann integrable, |f (x)|2 dx < ∞
0
n Z 1 o
2 2
LL ([0, 1]) = f : [0, 1] → C f is Lebesgue integrable, |f (x)| dx < ∞
0
2
as metric spaces with the metric
Z 1 1/2
2
d(f, g) = |f (x) − g(x)| dx
0
1
It is Theorem 4.15.
2
To be picky, d(f, g) isn’t a metric, because there are nonzero functions f with d(f, 0) = 0. This
is a technicality which is easy to deal with.
Introduction 4 September 4, 2019
Then L2R ([0, 1]) is not complete3 , while L2L ([0, 1]) is complete. L2 spaces are per-
vasive in quantum mechanics.
• The measure theory approach to integration is easy to generalize. For example
one can rigorously implement the “Dirac delta function”
( )
0 if x 6= 0
Z
δ(x) = with δ(x) dx = 1
+∞ if x = 0
by making a suitable definition of the “size of a set”. In fact, our framework will
be flexible to define an integral over any set, not just subsets of Rn .
3
See the notes “Incompleteness of L2 with the Riemann Integral” for an example of a Cauchy
sequence that does not converge.
Introduction 5 September 4, 2019
