Analysis I
Real Analysis 1
Math 623
Notes by Patrick Lei,
May 2020
Lectures by Robin Young, Fall 2018
University of Massachusetts, Amherst
Disclaimer
These notes are a May 2020 transcription of handwritten notes taken during lecture in Fall 2018. Any errors are mine and not the instructor’s. In addition, my notes are picture-free (but will include commutative diagrams) and are a mix of my mathematical style (omit lengthy computations, use category theory) and that of the instructor. If you find any errors, please contact me at plei@umass.edu.
1 Lebesgue Measure
1.1 Motivation for the Course
We will consider some examples that show us some fundamental problems in analysis.
Example 1.1 (Fourier Series).
Let be periodic on . Then if is sufficiently nice, we can write
where . If this is possible, then we have Parseval’s identity:
However, the function space we are working in on the RHS is not complete because there exist sequences such that the associated function is not Riemann-integrable. This tells us that we need a better notion of integration.
Example 1.2 (Dirichlet Function).
Enumerate the rationals in the interval by . Then define
Then the limit of is simply the indicator function . Here, the are Riemann integrable, but is not.
Next, we can consider the problem of characterizing which functions are Riemann-integrable.
Theorem 1.3.
A function is Riemann-integrable if and only if the set of points where is not continuous has measure .
Fourth, we may consider lengths of curves. Suppose we have a fractal (this will remain undefined) such as the Cantor set, a coastline, or broccoli. What is is the correct dimension for such a set? When is the “length” of such a curve finite?
Finally, recall the fundamental theorem of calculus: For a differentiable function ,
and for an integrable function ,
When do these two results apply?
Question 1.4.
All of these examples tell us that the fundamental problem in analysis is the exchange of limits. We will see that many of the results in the course are of this form.
1.2 Basic Notions
First recall the very basic notions of set theory: unions, intersections, complements, relations, and functions. We use the usual definitions of image, preimage, and partial and total orders.1 We will also assume the axiom of choice, so in particular, we have the following results:
- Hausdorff Maximal Principle:
-
Every partially ordered set has a maximal totally ordered subset.
- Zorn’s Lemma:
-
If is a partially ordered set where every chain has an upper bound, then has a maximal element.
- Well Ordering Principle:
-
Ever set admits a well-ordering.
Next, we will recall some basic properties of Euclidean space.
Theorem 1.5 (Bolzano-Weierstrass).
Every bounded sequence in has a limit point. Note that the lim sup is an accumulation point.
Next recall that with the Euclidean distance is a metric space. In fact, is a complete normed metric space, and the triangle inequality is implied by Cauchy-Schwarz. Recall that open balls are given by and closed balls are given by . Then a set is open if around any we can find an open ball and a set is closed if is open. We can then define the interior, closure, and boundary of a set in the usual way.
Example 1.6.
Consider . Then and , so is a countable dense subset of . Therefore, is separable.
Convergence is defined in the usual way.
Lemma 1.7.
Let be a metric space and . Then the following are equivalent:
- 1.
;
- 2.
For all , ;
- 3.
There exists a sequence such that .
1.3 Rectangles
We now build towards defining the Lebesgue measure. First, we call any set of the form
a closed rectangle. We define the volume of any rectangle by
It should be obvious what a cube and an open rectangle are.
Definition 1.8.
A collection of rectangles is almost disjoint if their interiors are disjoint.
Lemma 1.9.
If a rectangle is an almost disjoint union , then .
Now let small enough. Then , As , we obtain .
Corollary 1.10.
Let , where the union is not necessarily almost disjoint. Then .
Our goal is to extend this notion of volume to open sets in terms of rectangles.
Lemma 1.11.
In , any open set is a disjoint union of countable many open intervals.
Proof.
Let . Then there is a maximal interval . Then if , note that . Thus . However, each interval must contain a rational number, so there are only countably many of them. □
Therefore, if is open, we can declare its measure to be , where is a disjoint union of the intervals . Note that this is not convenient in higher dimension, so we use closed rectangles instead. Note that any open set in is a countable union of closed rectangles , but this union is not even almost disjoint.
Theorem 1.12.
In , any open set can be written as an almost disjoint union of closed rectangles. Given this, we expect .
Proof.
Grid by the points in . Then each cube of length satisfies either , or neither. At the first step, include all and exclude all . Then we reduce to and proceed by induction. □
Example 1.13 (Cantor Set).
This is not really an example of rectangles, but is an important example for this course. It is a closed, compact, totally disconnected set where every point is a limit point. Define the sets
- •
;
- •
;
- •
,
and then continue the process. Then we define the Cantor set to be .
First, note that is an intersection of closed sets, so it is closed. Then given , note that there exists such that , so are in different intervals in . Clearly there are no isolated points. Finally, note that is the set of all real numbers with no in their ternary expansions, so it is uncountable. However, note that is a disjoint union of open intervals with measure
Therefore, we expect .
1.4 Outer Measure
It is reasonable to expect that if is an almost disjoint union then . Instead of imposing this inequality for general , we will make it an inequality, so we drop the condition of being almost disjoint. Define for any the outer measure
We can show that directly. Note that is a union of closed subintervals of length , so we see that for all . Therefore, . For another example, it is easy to show that if is countable, then . To see this, let and then take the rectangle of width centered at . Then .
Properties of outer measure. Here are some properties of the outer measure:
- 1.
(Monotonicity) If , then . As a corollary, bounded sets have finite outer measure.
- 2.
(Countable subadditivity) If , then .
Proof.
Assume that each . Then for all , there exist countably many cubes such that , so we have
Then any , so
- 3.
For any , we can write
Proof.
It is easy to see that . In the other direction, assume is finite. Then , so . Choose open such that and .
Now and , so
- 4.
(Wannabe additivity) If and , then . Here .
Proof.
Choose and choose such that , , and . If , then . Therefore, there exist such that and , so
- 5.
If is a countable union of almost disjoint cubes, then .
Proof.
For each , pick such that . Then and are disjoint and a finite distance apart, so for each , we have , and therefore , and thus . □
Note we cannot conclude finite additivity for general disjoint sets .
1.5 Measurable Sets
Definition 1.14.
A subset is measurable if for all , there exists such that .
If is measurable, define its Lebesgue measure as . Here are some properties of measurable sets:
- 1.
Open sets are measurable;
- 2.
If , then is measurable;
- 3.
Countable unions of measurable sets are measurable;
- 4.
Closed sets are measurable;
- 5.
If is measurable, then so is ;
- 6.
Countable intersections of measurable sets are measurable.
This says that measurable sets form a -algebra. In addition, we should note that if is closed and is compact such that , then .2
Theorem 1.15.
If are measurable and disjoint, then .
Proof.
Assume that the are bounded. Then there exist closed with such that for all . Then for finite , is compact, so . Therefore uniformly in . Therefore, . The other inequality is from subadditivity.
Next, if some of the are unbounded, set to be the cube with sides . Then set and . Finally, set . Then each is bounded, so we have
Corollary 1.16.
- 1.
If with measurable, then ;
- 2.
If and some has finite measure, then .
Proof.
- 1.
Set . Then the are disjoint and measurable, so . Therefore,
- 2.
Set . Then . Then by assumption, so
Therefore .
□
Theorem 1.17.
If is measurable, then for all :
- 1.
There exists an open such that ;
- 2.
There exists closed such that ;
- 3.
If , then there exists compact such that ;
- 4.
If , then there exists such that .
Proof.
We only need to prove the last two, the first two are by definition.
- 3.
There exist a closed with . Then for all , set . Then the are measurable, and converge down to , so there exists such that .
- 4.
Note there exists containing such that . Then there exists such that , so we can set . Then
Corollary 1.18.
is measurable if and only if
- 1.
differs from a set by a null set;
- 2.
differs from an set by a null set.
Proof.
- 1.
For all there exists open such that and . Therefore, set . Clearly and by Corollary 1.16.
- 2.
The argument is the same.
□
Other Properties of Lebesgue Measure
- 1.
is translation invariant;
- 2.
If is measurable, then ;
- 3.
is reflection invariant.
Definition 1.19.
A set is a -algebra if it is closed under complements and countable unions.
The Borel Sets are the smallest -algebra containing the open balls.
Lemma 1.20.
If is measurable and , then the set of differences contains an open interval around .
Proof.
There exists an open set such that . Then is a disjoint union of open intervals. Set . Then we know that , so there exists such that . Then for small if and are disjoint, then . However, if , then is an interval. Thus . Finally, we see that , so for small enough . □
Theorem 1.21 (Vitali).
There is a non-measurable set .
Proof.
Enumerate the rationals in by . Now consider the set and choose one representative of each equivalence class using the axiom of choice.3 Call this set . Next, note that contains no interval because it contains no rationals. If , then , which would imply that . Thus must be non-measurable. □
Corollary 1.22.
Any set with positive outer measure contains a non-measurable set.
Proof.
Let be as in Theorem 1.21. Then let . At least one of these sets must be non-measurable. □
1.6 Measurable Functions
Our goal will be integration. We expect that if is a measurable set, then . More generally, we want
Functions of this form are called simple. We also want to have some results where the integral of a limit is equal to the limit of the integral. Similar to continuity, we can try to define to be measurable if is measurable for all measurable . However, this is too hard to check, we find a minimal condition.
Definition 1.23.
A function is measurable if for all , is measurable.
Here are some properties of measurable functions:
- 1.
is measurable if and only if is measurable for all open . Alternatively, is measurable for all closed .
- 2.
Continuous functions are measurable. In addition, if is measurable and finite-valued and is continuous, then is measurable.
- 3.
If is a sequence of measurable functions, then , and (if it exists) are all measurable.
- 4.
If are measurable, then are both measurable.
- 5.
Define almost everywhere if the set on which they differ has measure . If almost everywhere, then is measurable if and only if is measurable.
We will now prove some approximation theorems for measurable functions. First, we attempt to approximate them by simple functions.
Theorem 1.24.
If is measurable on , then there exists an increasing sequence of simple functions converging pointwise to .
Proof.
Define the sequence of functions
Clearly for all and . Now define the set
and the simple function
Clearly is increasing in . We also see that . Now set . Then we see that
for all . □
Corollary 1.25.
Suppose is measurable. Then there exists a sequence simple such that and .
Proof.
Write , where and . The rest is left to the reader. □
Theorem 1.26.
If is measurable, there exists a sequence of step functions such that almost everywhere.
Proof.
First, there exist simple converging to everywhere. Write . Then for all , there exists a finite collection of cubes such that . Therefore, there exist almost disjoint rectangles such that . Shrink the to obtain disjoint closed rectangles such that . Then there exist step functions and measurable sets such that
- 1.
;
- 2.
.
Then set , so has measure . Then for all , we have
Theorem 1.27 (Egorov).
Let be a sequence of measurable functions on with and pointwise. Then for all there exists a closed set with and such that uniformly on .
Proof.
For given , set
If we fix , then and . Therefore there exists such that . Now choose such that and set . Then
Choose such that . Then . To prove uniform convergence on , choose such that and . Then if , , so
Theorem 1.28 (Lusin).
Let be measurable and finite-valued on and . Then for all , there exists closed such that and is continuous.
Proof.
First, there exists a sequence of step functions such that pointwise. Then remove thin boundaries of rectangles to find such that and such that is continuous off of . By Egorov, there exists closed such that uniformly on and . Next, let , where . Now if , is continuous on and uniformly, so is continuous on . Finally, approximate by a closed set such that . □
Finally, we will construct a non-measurable function. First, we will define the Cantor function which is given by
Note that is monotone and continuous on and constant on .
Then define , so is a continuous bijection. On any excluded interval , sends to . Therefore, , so . Then there exists a non-measurable set . Then is not measurable.
2 Integration
2.1 Defining the Integral
Our goal is to integrate measurable functions. If is simple, then we define
To check that this is well-defined, if we can write , then we can consider . This gives a canonical representation of a simple function (where all of the are disjoint). It is easy to see that we have linearity, monotonicity, and the triangle inequality. We want to extend this to a larger class of functions.
Next, we will consider bounded measurable functions on sets of finite measure. If is supported on a set of finite measure and is bounded, we define
for some sequence of simple functions converging to . We need to show this is independent of the sequence.
Lemma 2.1.
If pointwise on with , and is bounded, then
- 1.
The limit exists;
- 2.
If almost everywhere, then .
Proof.
For the first part, we use Egorov. There exists a closed set such that uniformly on and . Let . Then
On , we have uniform convergence, so
which clearly converges to . Then the second terms are , which also converges to , where . Thus is Cauchy, so it converges.
If almost everywhere, then
so . □
By Lemma 2.1, integration is well-defined. Again, we have the desired properties. Before we extend the definition of the integral to all functions, we will prove a convergence theorem. This is the first exchange of limits result.
Theorem 2.2 (Bounded Convergence Theorem).
If is a sequence of functions, uniformly bounded by and uniformly supported on of finite measure, and it pointwise almost everywhere, then is measurable and . In fact, as .
Proof.
The proof of this is the same as proof of the consistency of the definition of the integral. □
Corollary 2.3.
If is bounded and supported on with , and if , then almost everywhere.
Proof.
Set . Then is bounded, supported on , and . Then
so . □
Theorem 2.4.
If is Riemann-integrable on , then it is Lebesgue integrable, and the integrals coincide.
Proof.
Translate the condition about partitions into step functions , where and . Clearly the are uniformly bounded. Therefore, by bounded convergence, are both measurable and bounded, and , so almost everywhere. Then clearly almost everywhere, so is measurable, and by monotonicity, . □
Finally, we can define integration for all nonnegative functions. We define
For any measurable , define . Note that it is easy to check that the definition makes sense. If , we say that is integrable.
Properties of the Integral.
- 1.
Linearity;
- 2.
Monotonicity;
- 3.
Additivity on disjoint sets;
- 4.
If is integrable and , then is integrable;
- 5.
If is integrable, then almost everywhere;
- 6.
If , then almost everywhere.
Note that if almost everywhere, then does not necessarily converge to . For example, if we take , then , but .
2.2 Some Convergence Results
We will state some convergence results where we can swap limits and integrals. We will also give the general definition of the integral here.
Lemma 2.5 (Fatou).
If is a sequence of functions with and if , then
Proof.
Suppose bounded satisfies and . Then set . Then is bounded, so by bounded convergence, . Also, , so . Because , we get the desired result. □
Corollary 2.6.
If with almost everywhere, then .
Proof.
By monotonicity, . □
Corollary 2.7 (Monotone Convergence).
If almost everywhere, then .
Corollary 2.8.
Given measurable, then . Moreover, if converges, converges almost everywhere.
Proof.
Let be the partial sums. Then by monotone convergence, if converges, also converges, so it is integrable. Thus converges almost everywhere. □
Example 2.9 (Borel-Cantelli).
Given measurable with , then . To see this, set and use the previous result.
Example 2.10.
Let
Then for all .
To see this, we will bound . Consider the annulus . Then set
Then set . Clearly , so is finite if is. However, we see that
We now give the general definition of the integral. Let be a measurable function and write , where both are nonnegative. Then we want . Clearly this is defined if at most one of the two integrals on the RHS is infinite.
Definition 2.11.
is integrable if both are finite.
Theorem 2.12.
The integrable is linear, additive, monotone, and satisfies the triangle inequality.
Lemma 2.13.
Let be integrable. Then
- 1.
Given , there exists a ball of finite measure such that ;
- 2.
For all , there exists such that whenever , then .
Proof.
- 1.
We assume . The for each consider and . Then we see that , so by monotone convergence, . Therefore, there exists such that
- 2.
Again assume . Then set and . Again, and is integrable, so there exists such that for all . Then choose such that . Now if , we see that
Theorem 2.14 (Dominated Convergence).
If almost everywhere and almost everywhere for some integrable , then is integrable and as .
Proof.
For , set . Then define . As in the lemma above, there exists such that .
Note that for all and is supported on , which has finite measure. Also, . By bounded convergence,
for large enough, so
We now briefly consider functions . Then note that , so . Also, because , . We conclude that is integrable if and only if are both integrable.
2.3 Vector Space of Integrable Functions
Let denote the set of integrable functions. Define the norm . Note that this is only a seminorm, so we set if almost everywhere.
Theorem 2.15.
is a normed vector space.
Theorem 2.16 (Riesz-Fischer).
is a Banach space.1
Proof.
Suppose is Cauchy in . Then for all choose such that . Then we define
Note that is nonnegative, so by monotone convergence,
so by dominated convergence, is integrable. In addition, , so we have a convergent subsequence. But then because the sequence is Cauchy and by the triangle inequality, it converges to . □
Corollary 2.17.
If in , there exists a subsequence such that almost everywhere.
Definition 2.18.
We say converges to in measure if .
Example 2.19.
Write natural numbers as , where . Set . Then this converges to in measure but not pointwise anywhere.
Theorem 2.20.
If in measure, there exists a subsequence almost everywhere.
Proof.
Given , there exists such that for . Then choose
Now if , then for . However, , so this is a set of measure . □
Remark 2.21.
In Fatou’s lemma, monotone convergence, and dominated convergence, we can replace pointwise convergence by convergence in measure.
Recall that in a topological space , a set is dense if for all open . In a normed space, this is equivalent to saying that for all and , there exists such that .
Theorem 2.22.
The following classes of functions are dense in :
- 1.
Simple functions;
- 2.
Step functions;
- 3.
Continuous functions of compact support.
Proof.
Take .
- 1.
We know that there exists a sequence of increasing simple functions pointwise, but is integrable, so .
- 2.
We approximate simple functions by step functions. Let be simple. Then there exist disjoint closed rectangles such that . This bounds .
- 3.
It is easy to approximate by continuous functions. Just build side ramps.
□
2.4 Symmetries
First, note that the integral is translation invariant. To see this, we can translate step functions. In addition, it is easy to see that .
Examples 2.23.
First, note that
Similarly,
Example 2.24 (Convolution).
Given integrable , define their convolution by
It is easy to see that .
Now write , so . Then for fixed write . Similarly, for , write . Define and similarly. The main issue is that if is measurable, then might not be.
Theorem 2.25 (Fubini).
Suppose is integrable on . Then
- 1.
For almost every , is integrable on ;
- 2.
The function is integrable on ;
- 3.
.
Proof.
Let be the set of integrable functions where all three conditions hold. We show that . First, we show that .
Given any rectangle , it is easy to see that . If we write , then . Then the first condition holds everywhere. For the second condition, note that
For the third condition, note that
Therefore, all step functions are in .
Now we will show that
- 1.
is closed under linear combinations;
To see this, let and . Then use linearity of the integral.
- 2.
is closed under limits.
We do this for monotone limits. Note that for all , there exists of measure such that is not integrable for . If and and is integrable, we want to show that . Let correspond to . Then their union has measure , so is integrable almost everywhere.
Now set . Now each is integrable and , so , so and thus almost everywhere. Thus is integrable almost everywhere, so .
Recall that if is integrable, there exists a set of simple functions converging up to . Each simple is a linear combination of for some measurable of finite measure, so we prove .
This reduces the problem to measurable sets:
- 3.
sets;
If is contained in the boundary of a rectangle, the result is obvious. Next, suppose is a finite union of almost disjoint rectangles . Then write as a linear combination of the and the , where . Thus .
Now if is open and of finite measure, write , where the are almost disjoint cubes. Then write , which is a monotone limit. Finally, if for open, then is also a monotone limit.
- 4.
Null sets;
Recall that there exists a set of measure such that . Then , so , so almost everywhere. Because for all , is measurable and null almost everywhere, so .
- 5.
Limiting procedure;
For any measurable , differs from a set by a null set.
Finally, for general , write , which are both integrable and nonnegative. □
Theorem 2.26 (Tonelli).
Fubini extends to measurable functions for which is defined but infinite.
Corollary 2.27 (Cavaleiri’s Principle).
Let be a measurable subset of . Then for almost every , the slice is measurable, and .
3 Differentiation
3.1 Integration and Differentiation
Recall the fundamental theorem of calculus:
- 1.
For continuous, the function is differentiable and .
- 2.
If is Riemann-integrable, then .
This roughly tells us that integration is anti-differentiation. We can replace differential equations with formulations in terms of integrals. Then recall that
What happens when we generalize to higher dimensions and consider ? Define . Note that need not be integrable.
Theorem 3.1.
- 1.
is measurable;
- 2.
almost everywhere;
- 3.
There exists depending only on such that for all ,
Proof.
- 1.
Define . Then note that if and only if
for some ball . In fact, for all , , so is open.
- 2.
This is a corollary of part 3.
- 3.
The proof of this relies on the Vitali Covering Lemma, which is stated below.
Let . Then we know if and only if there exists such that . Given a compact , write and extract a finite subcover. Then we use Vitali, so
Because is arbitrary, we have the desired result.
□
Lemma 3.2 (Vitali’s Covering Lemma).
Let be a finite set of open balls. Then there exists a set of disjoint open balls such that .
Proof.
Inductively choose to be the ball of largest radius not disjoint to thre previous balls and not contained in . We started with a finite collection, so the process terminates. □
Theorem 3.3 (Lebesgue Differentiation Theorem).
If is integrable, then for almost every ,
In , this implies .
Proof.
For , define the set
We will show that .
Let . Recall that continuous functions of compact support are dense in , so choose a continuous function with compact support such that . Then is uniformly continuous, so as . Now use the triangle inequality:
Therefore, we have
Let be the set where and where . By Theorem 3.1, , and by Chebyshev, . □
Corollary 3.4.
For almost every , .
Also note that we can replace integrability by local integrability in these results. If is measurable, define to be a point of density of if
Corollary 3.5.
Let be measurable. Then
- 1.
Almost every is a point of density;
- 2.
Almost every is not a point of density.
In a related notion, we can define the Lebesgue set of as the set of all such that and
Clearly if , then the Lebesgue differentation theorem holds. Also, if is continuous at , then .
Corollary 3.6.
For any , almost every belongs to .
Proof.
For rationals , consider . Then there exists with such that the Lebesgue differentiation theorem holds for for all . Thus if , then there exists such that for all . For this , choose such that if , then
so . Therefore
Note that any locally integrable function has different representatives up to sets of measure . We know that for almost every , and if almost everywhere, , but up to a set of measure .
On the Lebesgue set, we can generalize average. Instead of balls, we can consider general sets . Note that , which is morally equal to , where the kernel satisfies . We will actually consider families of such kernels .
3.2 Kernels
Consider sets for which converges.
Definition 3.7.
A family of sets shrinks regularly to if there exists such that for all , there exists such that , and .
Corollary 3.8.
If is locally integrable and shrinks regularly to , then
Proof.
Note that
Note that is measurable if and only if for all , . Also, note that . Instead of multiplication, we will convolce. We will show later that commutes with differentiation.
Note that is a restriction of . Far from , and near . However, is not usually differentiable. Instead, we will convolve with different kernels.
Definition 3.9.
A good kernel satisfies
- 1.
for all .
- 2.
uniformly. Also, for all ,
as .
Example 3.10.
We will construct a smooth bump function. Let
Now set . Then set . This is a smooth even function with integral supported on . Now we will introduce scaling, to define a family defined by
where is chosen such that . If we define , then is a smooth approximation to that is . In addition, the derivative of is .
The bottom line is that for all , is . Also, as for all Lebesgue points of .
We call a kernel an approximation to the identity if it satisfies
- 1.
;
- 2.
for all ;
- 3.
for all .
Therefore, as for . If , as . Note that the Dirac mass1 is a “generalized function” which satisfies , for , and .
Example 3.11 (Poisson Kernel).
Consider the upper half plane (or equivalently the disk). Then define
for . Then for , as . On the disk, define
Note that away from the origin. If we convolve , solves on and .
Example 3.12 (Heat Kernel).
Define the heat kernel
Again, if , as . If , then . Also, for , . Note that for , the heat equation with initial conditions is solved by .
Because the heat equation is linear, since is a convolution identity, commutes with , so given , set . Because differentiation commutes with , we see that
but at , then .
Example 3.13 (Fejer Kernel).
This kernal is defined by
It is related to the Fourier transform. Recall that the Fourier transform is defined by
and its inverse
Note that as a function of is a wave.
Theorem 3.14.
If is measurable and , then as .
Proof.
We calculate
We will show that . We will use properties to and split the integral into parts and . Define for ,
where . We then need Lemma 3.15, which is stated after this proof. This yields boundedness on . Then for large , we use the triangle inequality:
Turning back to our original problem, note that
The first term is given by
On the annulus, we will estimate . Then we can write
so
For large , , so for any , we see that
Therefore, given , choose such that . Then for all sufficiently small ,
Lemma 3.15.
If and , then is continuous on and as .
Proof.
Because the integral is absolutely continuous, is continuous: absolute continuity of the integral means that for all , there exists such that if , then .
Now consider small annuli. For fixed, the annuli have measure , which is small. Then is just the statement that . □
3.3 Differentiation
Let be monotone increasing. Given only two points , for any increasing sequence , we have
In particular, this series converges absolutely, so there are only countably many discontinuities, and moreover , where and is defined similarly.
We know that for all and that the three values are equal if and only if is continuous at . Then means that there are countably many jumps and that their sum is finite. Now assume that is continuous and increasing on .
Note that the Cantor function is continuous, monotone, and takes . Recall that is constant on each excluded interbal, so almost everywhere. Therefore, , but .
Theorem 3.16.
If is continuous and monotone on , then the derivative exists almost everywhere and
Before we prove the theorem, we need to define the Dini derivatives. Define
Then the Dini numbers are defined everywhere and that . Then is differentiable at if and only if and and all these values are finite.
Lemma 3.17 (Rising Sun Lemma).
Let be continuous on . Then the set is a relatively open union of disjoint intervals , and for all , .
Proof.
It is clear that the set is open. Now if , then we can extend the interval to the left, and similarly to the right if . □
Lemma 3.18 (Vitali’s Covering Lemma).
A collection of open/closed interbals is a Vitali cover of if for all and , there exists such that and . Let be of finite measure and a Vitali cover. Then for all , there exists a finite set of disjoint such that .
Proof.
Assume without loss of generality that each is closed. Fine of finite measure with . Now choose a series of intervals as follows. Suppose that have already been chosen. Then let
Either or there exists contained in and disjoint from the other sets such that . Continue the process. If this does not stop, then choose such that . Let .
To show that , choose . Because the are closed, there exists such that and . Thus does not intersect any of the . From our construction, there exists such that . however, tends to , so there exists a smallest such that is nonempty. For this , if is the midpoint of , we see that . However, , so . Therefore for some with , so
Proof of Theorem 3.16.
We want to show that almost everywhere and amost everywhere. Fix and set . We will show that is measurable. Note that is measurable and is continuous, so replace with . Set .
The idea is that on , . Then there exists such that for . Thus , so . is not monotone but is continuous. We will now eliminate and . Note that if and only if for some . By rising sun, we know that is open and of finite measure, so .
Now choose rational and set . We will show that . If not, then there exists an open with . On some connected component of , apply rising sun to on . Then we obtain intervals , so wet . Then we see that for .
To show that , find an open such that . Then for all , there exists an interval suchthat . This is a Vitali covering.
Now choose such that . Summing over this finite subcover, we see that
Then if we set to be the interior of , then for all there exists such that and . By Vitali, there exists of the form such that . For each , we have , so adding, we get that
Each is contained in an , so we can sum over to see that
Because is arbitrary and , we see that . □
We now claim that is measurable. In addition, is integrable and
To see this, note that is an almost everywhere limit of measurable functions. By Fatou, we see that
Because is continuous,
for large enough, so . The same argument holds for the integral near .
3.4 Bounded Variation
Given , create a finite partition . Then set
Then define the increasing, decreasing, and total variations of by
We say is of Bounded Variation if is finite.
Lemma 3.19.
If is BV, then and .
Proof.
Choose a finite partition. It is easy to see that , so for all partitions, . Then , but this means . Similarly, if we take the supremum over first, we get the inequality reversed. The first part of the lemma should be easy. □
Theorem 3.20.
A function is BV if and only if it is the difference of two bounded monotone increasing functions.
Proof.
If is BV, set and . Because is BV, are monotone increasing and bounded, and by Lemma 3.19, . On the other hand, if with monotone and bounded, then is finite. □
Corollary 3.21.
If is BV, then exists almost everywhere.
Now we will consider jumps. Assume is bounded and monotone increasing. At each point, the limits and (defined analogously) exist. By monotonicity, . Define the jump set to be the set . We know that is countable, so enumerate it by . For each , define . At each , we have for some . Now define
and define the jump function of by .
Lemma 3.22.
- 1.
is monotone increasing;
- 2.
is monotone increasing and continuous;
- 3.
is differentiable almost everywhere and almost everywhere.
Proof.
- 1.
This is obvious.
- 2.
The series defining converges uniformly and , so if , each is continuous at , so is continuous at . If is a jump, then . In particular, , and .
Now at . Also, provided that is increasing, because it is continuous. If , note that
so we see that is increasing.
- 3.
Note that is discontinuous on a countable set. Given any , there exists a closed such that and . Then is open and is constant on each component of . Therefore, the set where does not exist or is nonzero has measure less than for all . □
Morally, we say that , where is the Heaviside step function, and .
We now consider the question of which continuous monotone functions satisfy the Fundamental Theorem of Calculus. We consider properties of , where is integrable on . We will refer to the integral as a function .
Lemma 3.23.
is continuous and of bounded variation.
Proof.
Continuity is easy. Use the fact that for all , there exists such that if , then . To show that is BV, choose . Then we see that
which is finite. □
Lemma 3.24.
If is integrable and for all , then almost everywhere.
Proof.
Set , where and . If , then there exists closed with . Then is relatively open. Either we have or . Then , so write . Therefore, at least one of the is nonzero, so either or is nonzero. □
Theorem 3.25.
If is integrable and if , then almost everywhere.
Proof.
If is bounded by , then we know is BV, so exists almost everywhere. Let , so
Also, almost everywhere. By bounded convergence, we then have
so almost everywhere.
For the general case, assume that . Then set . Then , so is monotone increasing. Thus it has nonnegative derivative almost everywhere. By the above, we know that
almost everywhere. Therefore
almost everywhere. Thus , so , so then
3.5 Absolute Continuity
Definition 3.26.
A function is absolutely continuous if for all , there exists such that if for a finite set of non overlapping intervals and if , then
Note that absolute continuity implies uniform continuity. From the previous lemma, then integrability of means that is absolutely continuous.
Lemma 3.27.
If is absolutely continuous, then it is also BV.
Proof.
Let be absolutely continuous and choose . Then given any partition , choose . Add in many partition elements so that is covered by sets of subinterbals each of length less than . Now
Therefore . □
As a consequence, if is absolutely continuous, then its derivative exists almost everywhere.
Lemma 3.28.
If is absolutely continuous on and if almost everywhere, then is constant.
Proof.
Fix and set , so . Let two parameters be given: an upper bound for on and a bound for . Use absolute continuity on : Choose according to the definition of depending on .
If , then . Thus there exists such that and , so . Now use Vitali’s covering lemma, so there exists a finite non-overlapping set of interals and such that and
Order the and set . Now we have
Theorem 3.29 (Fundamental Theorem of Calculus, Part 2).
A function has integrable such that
if and only if is absolutely continuous.
Proof.
Suppose that is absolutely continuous. Then is BV, so for monotone increasing . Therefore exists almost everywhere and . Therefore
Set and . Then is also absolutely continuous and almost everywhere from the definition of , so almost everywhere. Thus is constant and must be equal to . □
We will now characterize functions of bounded variation.
- 1.
If is BV, then .
- 2.
has a jump part , where .
- 3.
is continuous and the difference of two monotone continuous functions. Thus exists almost everywhere. If we set
then is absolutely continuous. Then the function is monotone increasing, but almost everywhere, so it does not satisfy FTC (behaves like the Cantor function).
In conclusion, we can write .
4 General Measures
Let be a set and be a -algebra of . Then is a measure on if is a nonnegatie -additive function on .
Examples 4.1.
- 1.
The Lebesgue measure as defined previously;
- 2.
The counting measure;
- 3.
The discrete measure on where each element is assigned a weight;
- 4.
Let be an uncountable set, set to be the countable or cocountable sets, and then let when is countable and otherwise.
Lemma 4.2.
The following properties hold for all measures:
- 1.
Monotonicity;
- 2.
Countable subadditivity;
- 3.
If and , then if the are nested.
Proof.
- 1.
Observe that and then use additivity and nonnegativity.
- 2.
Let . Then .
- 3.
Note that , so
We say that is finite if and that is -finite if is a countable union of sets of finite measure.
Example 4.3.
- 1.
The Lebesgue measure is -finite (just take for ).
- 2.
The counting measure is finite if and only if is finite.
- 3.
Then discrete measure if finite if and only if countably many points have positive weights and the sum of the weights is finite. It is -finite if and only if only countably many points have positive weight and each weight is finite.
- 4.
This is a finite measure. In fact, any measure such that is a probability measure.
4.1 Outer Measures
An outer measure is a countable subadditive function defined on such that . Our main question is to extract a measure from an outer measure when we have no recourse to open sets, unlike in the Lebesgue case.
Definition 4.4.
is (Caratheodory)-measurable if for all , .
Note that
- 1.
This condition is equivalent to ;
- 2.
The condition holds for the emptyset;
- 3.
is measurable if and only if is.
This tells us that is a -algebra and that is a measure on .
To prove the last claim, we need to show countable additivity. First, we will do it for finite unions:
Also, additivity on disjoint sets is easy. Now we will consider countable disjoint unions. Set and . Then
Therefore, , so , so if we allow , then
Now let be a metric space. As before, the Borel sets are the smallest -algebra containing all open balls. An outer measure is a metric outer measure if it has the separation property.
Theorem 4.5.
If is a metric outer measure, then the Borel sets are measurable and is a measure.
Proof.
We show that closed sets are measurable. Assume is closed and . Then set
Then clearly the are increasing. We will show that .
Set . Then . Then we see that
Then we see that and , so in particular, . Then we obtain
Because the sum is finite, we have the desired result. □
Theorem 4.6.
Let be a Borel measure that is finite on compact sets. Then for every and , there exists an open and closed such that and .
Proof.
Let be a collection of sets satisfying both properties. We show this is a -algebra containing . Clearly, it is closed under complements. Thus suppose for each . First, there exist open such that , so their union is open and .
Similarly, there exists suchthat , but the union is not closed. We prove that there exists a closed such that . Assume that are increasing, so choose and set . Then let , so write , where .
For each , there exists such that with . Then we see that , so and each is closed, so is closed.
Finally, we show that open . Set . Then we use the argument above. □
4.2 Results and Applications
Definition 4.7.
A premeasure is a measure-like function on an algebra (closed under complement and finite union).
We see that a premeasure defines an outer measure. Define to be measurable if the preimage of any open set is a Borel set and define to be simple if
for measurable. Then given a measure , definee
where .
Theorem 4.8.
is a measure where is the smallest -algebra containing .
We have the following results, in analogy with the Lebesgue case.
Theorem 4.9 (Egorov).
Let be defined on with . Then for all , there exists with and such that uniformly on .
Lemma 4.10 (Fatou).
If is nonnegative, then .
As before, if with both , then is defined provided that not both and are infinite. Then is
- 1.
Integrable if . We say ;
- 2.
Square-integrable if . We say that .
Theorem 4.11 (Monotone Convergence).
If and almost everywhere, then .
Theorem 4.12 (Dominated Convergence).
If almost everywhere and for some integrable , then .
Example 4.13.
Consider the Stieltjes integral. Recall that the Riemann integral is , and the Stieltjes integral is for some monotone . Finally, the Lebesgue integral is the correct definition of integration.
To construct a measure associated to a monotone function , first normalize by changing at the jumps to ensure right continuity.
Theorem 4.14.
Given a monotone, right-continuous , there exists a unique Borel measure such that for all . Moreover, if is a Borel measure that is finite on compact sets, there exists a unique (up to constants) such that .
Proof.
The first part is straightforward, but painful (Robin’s words), so it is not given here. On the other hand, if is a given Borel measure and is finite on compact sets, how do we define ? We want . If , then we must have and similarly for . It is easy to see that this function is monotone, so we show right continuity:
□
4.3 Signed Measures
Note that we can add measures and multiply them by nonnegative constants. How can we obtain full linearity?
Definition 4.15.
A function on a -algebra is a signed measure if
- 1.
;
- 2.
does not take on both ;
- 3.
is additive on disjoint sets. In particular, if is finite and , then converges absolutely.
We say that is positive with respect to if and for each with , we have . Define similar for negativity. Then a set is null if it is both positive and negative. However, having measure does not imply that a set is null.
Lemma 4.16.
Subsets and countable unions of positive sets are positive.
Proof.
Subsets of positive sets are positive by definition. For the second part, if , set . Then and . □
Lemma 4.17.
If satisfies , then there exists a positive with .
Proof.
If is positive, we are done. Otherwise, let be the smallest integer such that there exists with . Inductively, choose to be the smalles integer such that there exists with . This process may or may not terminate, so set . Then clearly has positive measure because and converges absolutely. Therefore is finite, so either the process stops or , so .
To show that is positive, it is clear if the process stops. Otherwise, for all , there exists such that . Then , so contains no subset with measure less than . □
Theorem 4.18 (Hahn Decomposition Theorem).
Given a signed measure , there exists positive and negative such that and .
Proof.
Assume that does not take the value . Define . Note that . Then choose positive such that and set is positive and that .
Also, , so each . Thus for all , so . Set . We will show that is negative. To see this, if with , then is positive with measure strictly greater than . □
Note that this are not unique. Then define the positive and negative parts of by and . These are both measures. Then we see that . This decomposition is unique. Also, define the total variation of as .
Note that is defined because cannot both be infinite. If , we say that is finite. If two measures have sets such that and , then we call them mutually singular and write .
Example 4.19.
The discrete measure is mutually singular with the Lebesgue measure. The Cantor measure is also mutually singular with the Lebesgue measure.
The opposite notion is absolute continuity of measures.
Definition 4.20.
We say that is absolutely continuous with respect to , written , if for all with , .
Example 4.21.
If is measurable and is integrable, then is absolutely continuous with respect to .
Theorem 4.22 (Radon-Nikodym).
Let be a measure space with Borel sets and suppose is -finite and . Then there exists a nonnegative -measurable such that for all , . Here, is called the Radon-Nikodym derivative of , sometimes written .
Before we prove this result (following Royden), we need two lemmas:
Lemma 4.23.
If is a collection of Borel sets ordered by some countable satisfying for , then there exists a measure such that on and on .
Proof.
Given , set . Then if , . Then if , because for . To show that is measurable, note that
which is measurable. □
Lemma 4.24.
Let countable and such that if , then . Then the same conclusion as in Lemma 4.23 holds.
Proof.
Let . Set . Note that has measure . Then apply Lemma 4.23 to the . □
Proof of Radon-Nikodym.
Assume is finite. For all rational , consider the signed measure . By the Hahn decomposition theorem, there exist with on and on . Then set and note that , so on and on . Therefore, if , then , so by Lemma 4.24, there exists such that on and on . Because , .
We now show that . Fix and set
Then is the disjoint union , so
Now , so on , so
By definition of , we know that , so
Then add and take . □