Analysis II
Abstract
Continuation of Math 623. Introduction to functional analysis; elementary theory of Hilbert and Banach spaces; functional analytic properties of Lp-spaces, applications to Fourier series and integrals; interplay between topology, and measure, Stone-Weierstrass theorem, Riesz representation theorem. Further topics depending on instructor.
1 Organizational
A webpage (http://people.math.umass.edu/~young/analysis/index.html) has been put up with one topic per week, which is impractical. We will aim to get through ten topics. Formal office hours will not be scheduled because this is the only class Robin is teaching. The understanding is that we will visit his office. At some point during lecture, the electrical engineering student walked into the room anad had the following conversation with Robin:
- Robin
-
Are you Bo?
- Bo
-
Yes
- Robin
-
Everyone, this is Bo. Bo, this is everyone.
- Bo
-
I am in the electrical engineering department. This will help me with my research.
2 Review of Last Semester
2.1 Lecture 1 (Jan 22)
We will review measure theory from an abstract perspective and treat the Lebesgue measure as an example.1 Let be a locally compact topological space.
Definition 1 (Sigma-Algebra).
A -algebra to be a collection of subsets closed under complement and countable union.
We denote by the Borel -algebra, which is the smallest such containing all open sets. is a (non-negative) measure if is defined on , the empty set has measure , and is countably additive.
Remark 2 (Notation).
Indexing with a Roman letter means countable and indexing with a Greek letter means uncountable.
Definition 3 (Sigma-finite).
A measure is -finite if where each has finite measure.
Definition 4 (Completeness).
A measure space is complete if all subsets of measure zero sets are measurable: and implies .
Definition 5 (Measurable function).
A function is measurable if is measurable for all real .
Given the notion of a measurable function, we can integrate:
Definition 6 (Simple Function).
A simple function is a finite linear combination of indicator functions
We define the integral of a simple function to be
Given a nonnegative measurable function , we define the integral
where ranges over all nonnegative simple functions less than or equal to almost everywhere.
We also set, for a measurable set ,
Example 7.
Given the trivial sigma-algebra, only constant functions are simple.
Lemma 8.
Any measurable can be approximated from below by with .
We observe that Fatou’s Lemma, monotone convergence, and dominated convergence hold.
Lemma 9 (Convergence theorems).
If setwise on and if almost everywhere (with respect to ) on , then
The ability to exchange limits and integrals is the motivation behind measure theory, and more broadly, analysis is about exchanging limits.
Definition 10 (Mutually Singular).
Two measures are mutually singular, written if there exist disjoint such that such that .
Example 11 (Cantor).
Consider the Cantor set and Cantor function . Recall that the Cantor function is a continuous surjection from to itself and has derivative zero almost everywhere. In particular, violates the fundamental theorem of calculus. Define the Cantor measure on the Borel sets in the usual way by . We see that , but where is the Lebesgue measure. Therefore, on .
Definition 12 (Absolute Continuity).
Given two measure , is absolutely continuous () with respect to if implies that .
Example 13.
Given any nonnegative integrable with respect to , we can define a measure . Then .
Theorem 14 (Radon-Nikodym).
If on a -finite (with respect to ), then there exists a non-negative measurable such that . We write .
Example 15.
Let be monotone increasing. Recall that left and right limits exist everywhere. Then observe that has at most countably many discontinuities and
for restricted to a compact set. We can write where encodes the jumps and is continuous. Then decomposes into , where is absolutely continuous with respect to the Lebesgue measure, and is the “Cantor” part. To extract we will consider the points at which is differentiable. In the end, we will have , where is absolutely continuous, is “Cantor,” and denotes the jumps. The three induced measures will be mutually singular.
2.2 Lecture 2 (Jan 24)
2.2.1 Inanities
Robin came in late and we had a conversation about Hagoromo Fulltouch chalk and chalk holders. Apparently those are like mechanical pencils. Robin would like to continue with the review and do a calculation.
2.2.2 Review Continued
Recall that we have a monotone increasing function , which we will assume is right-continuous. Recall that we can decompose as we did yesterday. There are several questions:
- 1.
How do we find this decomposition?
- 2.
What do we do with it?
To find , the jump part, we can locate jumps as the discontinuities. Let be the set of jumps. Then
where is the Heaviside function. Then we write the measure
where is the Dirac mass.
Now to find , we differentiate . Recall that we can compute the Dini derivatives everywhere. We also proved in the fall that the set of points of nondifferentiability has measure zero with respect to Lebesgue. Then, we can integrate with respect to to get
We know that , so is the absolutely continuous part of . Finally, the “Cantor” part is the remainder .
The bottom line is that it is hard to locate the “Cantor” part of a function.
Example 16.
Suppose with jumpts at . Then we see that We can differentiate to find . Then, integrating, we find that has no “Cantor” part. Now we see that the Lebesgue-Stieltjes measure is and integrating some function is something like
Conversely, given a measure , we can define the corresponding function satisfying . Define by for positive and similarly for negative
Recall again the Radon-Nikodym Theorem:
Theorem 17 (Radon-Nikodym).
Let be a -finite measure space and another -finite signed measure on . Then there exist unique measures , where , such that . Moreover, there exists integrable such that for all measurable . We write the Radon-Nikodym derivative.
Now recall signed measures:
Definition 18 (Signed Measure).
A signed measure is a countably additive function on that vanishes on the empty set and does not take on both positive and negative infinity.
Lemma 19.
There exist such that for all and for all .
We can set and . Then we have and . In addition, the decomposition is unique. Similarly, we can define complex measures. Then is the “total variation” of .
3 Product Measures and Fubini
3.1 Lecture 2 continued
Our first issue: are products of measurable sets measurable? What is the correct product -algebra? We will simplify the structure.
Definition 20 (Semi-algebra).
A semi-algebra satisfies: if , then and can be written as a finite disjoint union , where .
This is very convenient for Cartesian products because and .
Generate an algebra from our semi-algebra by plus finite unions of elements of . Now suppose we have an additive set function defined on , this extends to a measure on provided that:
- 1.
If is a finite union of disjoint union of elements of , then .
- 2.
is countably subadditive on .
Given and , we develop the product measure on . First let . Obviously, is a semialgebra, so it extends to an algebra . Now define on by .
Lemma 21.
satisfies the above two conditions.
3.2 Lecture 3 (1/29)
3.2.1 An Apology
Robin would like to offer an apology. He realizes that this is the first time he’s teaching the class, so we are all guinea pige here. He has come to realize that Stein’s book is not very good. It’s boring, encyclopedic, and not very well arranged. It has too much measure theory, and he apologizes to us for spending too much time in the fall on measure and integration, which is really boring. After we summarize Fubini in a general setting, we will take measure theory for granted and move on to something more interesting.
Robin says that the cardinal rule of teaching is to never end on time.
3.2.2 Product Measures/Fubini Continued
Recall the definition of a semi-algebra from last time.
We restate the lemma from last time:
Lemma 22.
Suppose we have an additive set function defined on , this extends uniquely to a measure on provided that:
- 1.
If is a finite union of disjoint union of elements of , then .
- 2.
is countably subadditive on .
Proof.
If , write each as , so . For the second part, we show that . We see that for all . Then we see that
so letting , we have our desired result. □
Now we return to our product construction from last time.
Lemma 23.
satisfies the above two conditions.
Proof.
First we show that if is a countable disjoint collection with , then . To see this, fix . For all , we see that is in only one , so , so as a function on , by countable additivity. Integrating with respect to and using monotone convergence, we have
and therefore . □
However, to achieve our goal, we need to extend this algebra to a -algebra so we can take limits. In order to do this, we need to perform Caratheodory’s construction and use outer measures:
Then we declare to be measurable if for all . Then conclude that thereis a complete -algebra. on which the product measure is defined.
Now we integrate with respect to . In order to prove Fubini, as before we denote the slice .
Lemma 24.
is measurable if , where , and . In addition, is a -algebra containing , and can be defined on the completion of .
Lemma 25.
If and is finite, then defined by is measurable and .
Lemma 26.
If , then for almost every , .
Theorem 27.
Let be a measurable subset of with finite. Then for almost every , is measurable and is a measurable function and .
Theorem 28 (Fubini).
- 1.
defined by is integrable on for almost every . The same statement holds if we switch the roles of and .
- 2.
is an integrable function on . Again, the same statement holds for .
- 3.
.
Theorem 29 (Tonelli).
The same conditions in Fubini’s theorem hold for non-negative measurable functions on .
Example 30 (Counterexample if non-negativity is not required).
Consider (see the homework).
We are going to assume now that integration over a general measure has been mastered. After this, we will talk about convergence of functions. Because spaces of functions are infinite-dimensional, we don’t have nice convergence results that exist in the finite-dimensional case. We will see that the most important norms are , which correspond to integration, Euclidean, and uniform, respectively. We will treat all of our Banach spaces abstractly. “Abstraction makes things easier generally. However, when applying this stuff, we have to remove ourselves from this abstruct into real stuff.”
“We stopped early and banked three minutes so I don’t have to stop early again.”
4 Notions of Convergence
4.1 Lecture 4 (Jan 31)
In undergrad analysis and last semester, we always think about . Now, we’re going to always think about infinite-dimensional spaces.
Recall that in , we can have sequences This is really the only form of convergence in a finite dimensional space. Functions can converge in several ways:
- Pointwise
-
This is easy but not very helpful. Functions are only defined up to sets of measure zero. This is essentially convergence in at each .
- Uniform
-
This is much better than pointwise convergence. For example, a uniform limit of uniformly continuous is uniformly continuous. This is the same thing as convergence in the norm .
- Convergence in Measure, , etc.
-
These were considered last semester.
We will consider abstract convergence in metric spaces. Let be a metric space. Then inherits a topology given by the metric.
Lemma 31.
Let be a metric space and . Then is closed if and only if every convergent sequence in has a limit in .
This gives us a characterization of closure using sequences.
Example 32 (Continuity).
Continuity on a metric space can be defined the same way as continuity in calculus. Alternative equivalent definitions include sequential continuity ( is continuous if it commutes with limits of sequences)and topological continuity. The first one is the most tangible and connected to , the second is the easiest to use in practice (but highly dependent on the notion of completion), and the third is the most abstract.
Definition 33 (Density).
is dense in if .
Remark 34.
If is dense, then for all there exists a sequence in converging to . Typically, if we can establish a property on , we can extend to by passing to the limit.
Example 35.
is dense in . We saw this last semester when we discussed convolution. Recall that if is a nice kernel and , then and the convolutions converge to . Thus, if we can define an operator on smooth functions, we can pass to the limit to define the operator on a much bigger class.
Example 36 (Heat equation).
The heat equation is . This “looks like” where and . So we’d like to write where is the Laplacian. The operator doesn’t make sense for general functions , but we can define it on using the power series. Then we can extend this to by taking limits.
4.1.1 Completeness
This is the same as the topological completion of to give . Recall that there are two constructions: Dedekind cuts, which are easier, and Cauchy sequences, which are better.2 Here, we will use Cauchy sequences which can be done in any metric space.
Definition 37.
is a completion of if:
- 1.
There exists an isometry ;
- 2.
is dense in ;
- 3.
is complete.
Example 39 (Equivalence of metrics).
Let be a positive definite symmetric matrix. Then is a metric with the “unit ball” the ellipsoid with axes given by . Then we see that is equivalent to the Euclidean metric, but not isometric.
Remark 40.
All isometries are injective. An invertible isometry is an isomorphism in the category of metric spaces.
Proof of Theorem 38.
Suppose we have two completions of . Then we find an isomorphism between .
Because is dense in , we use the definition of on to derive . Given any , there exists a sequence in where . Then we define .
We see that clearly this is well-defined because two sequences that converge to the same point must approach each other. To check that is an isometry, . Also, note that for all . Finally, note that if , choose such that . Then satisfies .
Existence is seen using Cauchy sequences. □
5 Compactness
5.1 Lecture 5 (Feb 5)
The notion of compactness distinguishes the two cases of finite and infinite dimensions. There are two notions of compactness: sequential and topological. A prototypical example of a compact set in is the closed interval .
Definition 41 (Sequentially Compact).
Let be a metric space. is sequentially compact if every sequence has a convergent subsequence with limit in .
Theorem 42 (Heine-Borel).
is sequentially compact if and only if it is closed and bounded.
This follows easily from the Bolzano-Weierstrass theorem, which states that every bounded sequence in has a convergent subsequence. Robin explained the idea of , but I will assume that you already know it. In the context of a metric space, we introduce the notion of “total boundedness.”
Definition 43 (-net).
Given a set , an -net is a subset of points of such that
Definition 44 (Totally Bounded).
A set is totally bounded if and onlg if has a finite -net for all .
Lemma 45.
If is totally bounded, then it is separable.
Proof.
For all , we extract a finite -net . Now let , which is countable. We show that is dense in .
Given , for any , there exists , but then there exists such that . □
Theorem 46.
is sequentially compact if and only if it is complete and totally bounded.5 Cauchy subsequence.
Proof.
First assume that is complete and totally bounded. Then given any sequence we extract a subsequence converging in . By completeness, we extract a Cauchy subsequence.
Choose and suppose inductively we have the subsequence chosen at the -th step. We want the subsequence to be Cauchy. Each subsequence is totally bounded. At the -th inductive step, set . Then the set is totally bounded, so has a finite -net. Then there exists a finite -net such that . Thus there exists one containing infinitely many . Choose and which are in . Clearly is Cauchy by construction.
For the converse, let be Cauchy in . Then there exists a convergent subsequence . The limit of any subsequence is the limit of the full sequence because the sequence is Cauchy, so is complete.
Now assume is not totally bounded. We find a sequence with no convergent subsequence. Then there exists such that has no fintie -net, i.e.for any finite -net, there exists not within of the net. For each , having chosen , choose such that but . By construction, for all , , so this sequence has no Cauchy subsequence. □
Definition 47 ((Topological) Compactness).
is compact if every open cover has a finite subcover.
In a metric space, compactness and sequential compactness are equivalent. However, there are examples of compact sets not sequentially compact and sequentially compact sets not compact.
Theorem 48.
Let be a metric space. is compact if and only if it is sequentially compact.
Proof.
Let be sequentially compact. Given an open cover of , we wish to extract a finite subcover.
First we use total boundedness to extract a countable subcover. We know is totally bounded, so it is separable. Let be a countable dense subset of . Let , which is countable. Then let . We use to extract a countable subcover of . Let . Then there exists such that . Then there exists such that . Now there exists such that . Also, . Now choose such that . Then . Thus is a countable open cover of . For any , there exists such that . These form a countable subcover of . We relabel them as .
Suppose there exists no finite subcover of . Then for all , there exists . Choose . Then there exists such that , but there exists such that . There must exists such that . Inductively, choose a sequence and a strictly increasing sequence of integers. For all , there exists such that whenever , . Therefore if , there is no subsequence . But the cover , has no subsequence converging to any , so is not sequentially compact.
In the other direction, suppose is not sequentially compact. Then there exists a sequence containing no convergent subsequence. Without loss of generality, suppose the are distinct. Then let . If cpmtaoms an for all , then that subsequence converges to . Thus there exists such that has no element of or for some . But then is an open cover. Thus there exists a finite subcover, which implies that there are only finitely many distinct elements of the sequence, which is a contradiction. □
5.2 Lecture 6 (Feb 7)
Robin actually arrived to lecture early today. This has never happened before in the history of this analysis class with the exception of the first lecture last semester.
Today we will carry on with compact sets and prove some theorems. One remark is that if you lose a derivative, you gain compactness.
Theorem 49.
Let is continuous with compact and a metric space. Then is also compact.
Proof.
Let be an open cover of . We pull back to an open cover of , extract a finite subcover, and then send this open cover back to the image. □
Theorem 50.
If from a compact set is continuous, then it is uniformly continuous.
Proof.
There exists such that for any , we can find such that , but . Now extract convergent subsequences and . Then , , but , which contradicts continuity because . □
Theorem 51.
Let be continuous on a compact . Then attains a maximum and minimum.
Proof.
Let . Then for all , there exists such that . Extract a convergent subsequence . By continuity . □
Corollary 52.
If is upper semi-continuous (resp. lower semi-continuous), then it is bounded above (resp. below) and attains its maximum (resp. minumum).
Remark 53.
The theorem says nothing about the uniqueness of optimizing sequences. An example is given by . Then there are concentric circles of minima.
5.2.1 Compactness in Infinite Dimensions
We introduce our metric space as the set of continuous real-valued functions on compact . Use the norm . Let denote the set of bounded functions . Then is a Banach space (a complete normed vector space). In fact is a Banach algebra: .
Theorem 54.
If is compact, then is complete and therefore Banach.
Idea of Proof.
Given a Cauchy sequence in , define the limit pointwise. Then each is uniformly continuous and is uniformly Cauchy (so converges uniformly to ), so . □
Remark 55.
In general, the norm is not a norm on but is a norm on .
We need compactness to obtain completeness. Define be the set of continuous functions with compact support.
Theorem 56.
Let be the closure of in . This is a Banach space.
Sketch of Proof.
Given a Cauchy sequence in , we need to show is in . We use a diagonal argument. Each is in , so each where each has compact support. Then we see that each . □
We sau that are the functions that “vanish at infinity.” Consider the prototype is .
Theorem 57 (Weierstrass approximation theorem).
Any function can be uniformly approximated by a polynomial. More precisely, the set of polynomials is dense in .
Proof.
Without loss of generality, take . Recall the binomial theorem . In particular for all . Thus . We then find the maximum of , which is . Therefore, we define .
We show that as . Note . Choose . Because is uniformly continuous, there exists such that implies . Set and .
Then
On , , so
Because has been fixed, on , , or . Therefore
which is at most
Therefore,
Now we compute . Note that . Differentiate with respect to and get . Therefore . Therefore
Again, we now get . We attain that . Multiplying through by , we obtain that . Therefore,
We conclude that . □
Lecture 7 (Feb 12)
When Robin walked into lecture, Jared asked him if he knew what a circumgalactic medium was, and Robin said it sounds like the aether. Robin has also posted homework, which is due a week from today. We will finish the proof of the Weierstrass approximation theorem, which is in the notes for last time.
Recall that in , is compact if and only if it is closed and bounded.
Definition 58 (Precompact).
is precompact if its closure is compact.
Note that this means is precompact if and only if it is bounded.
However, observe that is infinite-dimensional, so closed and bounded is not enough. Examples will come later. The concept that provides compactness in is equicontinuity. This is an analogue of uniformity across the functions.
Definition 59 (Equicontinuity).
Let be a set of functions . Then is equicontinuous if given any and , there exists such that if , then for all .
Definition 60 (Uniform Equicontinuity).
is uniformly equicontinuous if is independent of (Think the definition of uniform continuity or uniform convergence).
Theorem 61.
If is equicontinuous on where is compact, and , then is uniformly equicontinuous.6
Proof.
Suppose not. Then there exists such that for all , there exist such that , but . The goal is to show that is not equicontinuous.
We know that is compact, so from we can extract . Then the also converge to .7 However, because for all , we must have either or . This contracts equicontinuity. □
Theorem 62 (Arzela-Ascoli (1880s?)).
Let be compact. A subset is compact if and only if it is closed, bounded, and equicontinuous.
Proof.
There are three steps:
- 1.
If is unbounded, then it is not precompact.
- 2.
If is precompact, then it is equicontinuous.
- 3.
If is bounded and equicontinuous, then it is precompact.
Now we prove each part:
- 1.
This one is easy. Build a sequence such that . This sequence is not Cauchy, so there is no convergent subsequence.
- 2.
Assume is precompact. Then write . We can extract a finite subcover . Because each is uniformly continuous, then there exists such that if , then for all . Choose . Now if , then for any , choose such that . Therefore, .
- 3.
Let be bounded and equicontinuous. We will show that it is sequentially pre-compact. Given a bounded sequence, extract a Cauchy subsequence. Because is complete, this implies sequential compactness. Let be a bounded sequence. Because is compact, it has a countable dense subset .
Extract a subsequence of such that converges. Then extract a subsequence of such that converges. Then we extract a subsequence of such that converges. Continue this process to infinity. Now extract the diagonal sequence of functions . We know that at each , is Cauchy. We will show that is uniformly Cauchy.
We will use equicontinuity of . Because is compact, is uniformly equicontinuous. Thus there exists such that if , then for all . Because is compact with dense, . We extract a finite subcover . For each , the sequence is Cauchy. Therefore, there exists such that if , then . Choose . Then given any , for , choose such that . Then . □
5.3 Lecture 8 (Feb 14 )
Last time we saw that equicontinuity and boundedness implies precompactness, so what is equicontinuity?
Example 63 (Not equicontinuous).
Let be the skew hat function, which is piecesise linear and continuous with . Notice that the are not equicontinuous because for all . Therefore this sequence has no convergent subsequence, so it is not precompact. However, this sequence is clearly bounded.
Example 64.
If is a uniformly Lipshitz family, then it is equicontinuous. is uniformly Lipshitz if there exists such that for all , .
We can weaken this statement, but not by much.
Example 65.
Suppose our are represented by power series of the form . Then . Recalling that , we see that
Then note that can only be bounded by if . To use the factorization, the only possible bound is
This morally looks like , so the moral is that we need some sort of derivative to claim precompactness.
Remark 66.
We will see later that functions with derivatives are precompactly contained in functions with derivatives. Thus we can find a convergent subsequence of a bounded sequence, but the limit can only be expected to be .
5.3.1 Differential Equations
Example 67 (Existence theorem for ODEs).
Consider the ODE with initial condition . We will take and . We will explore the existence of solutions ot this equation. Assume that is continuous.
Recall Euler’s method, which is to set and approximate the integral term, taking steps of constant size. Given , we will build and then . Then we take to be the piecewise linear interpolation of these points.
Now we extract a convergent subsequence from the . We call an -approximate solution to the initial value problem if
- 1.
;
- 2.
is continuous, and it is differentiable at all but finitely many points of ;
- 3.
If exists at , then .
Then for , we have . We will use continuity of to make this small:
Theorem 68.
There exists an interval and a continuously differentiable function such that the IVP holds in .
Note that the above is not a global existence theorem. Let be our initial guess for the time of existence. Choose such that for each with satisfies . If , then set . Call . Then is compact, so is uniformly continuous in . Thus is uniforly Lipshitz continuous. Therefore is equicontinuous and bounded. By Arzela-Ascoli, we extract a convergent subsequence. Let . We know that .
We need to show that the limit is continuously differentiable and satisfies the IVP. We know that satisfies
Now we see that , which approaches by construction. Therefore . Because is continuous, then is differentiable and thus .
We have the following theorem for these solutions.
Theorem 69.
Let and suppose on . If . Then solution is defined ans satisfies for all and if is Lipshitz in and uniform in , then the solution is unique.
Example 70.
Consider the initial value problem , where . Then is a solution. Also we see that , so separating variables and integrating yields , which satisfies the initial condition. Therefore uniqueness fails.
The first part of Theorem 69 is the statement of time of existence. While remains continuous, the only way for solutions to fail to exist is for them to “escape to ,” or to leave any compact set.
5.3.2 Digression
Consider a -body problem. We consider gravitational attraction. We want to find out where solutions do not exist, where a solution is a continuous trajectory for each planet. The question is: how do solutions fail to be globally defined?
- 1.
Collision: f has a singularity and loses continuity. For the -body, problem, this is all;
- 2.
For the -body problem, there are initial conditions for which solutions escape to in finite time!
Example 71.
Consider a -body problem with two pairs of two overlapping elliptical orbits. Then the fifth body is an oscillator between the two pairs of elliptical orbits.
5.4 Lecture 9 (Feb 21)
Last time we discussed Peano’s proof of the existence of solutions to first order differential equations using the Euler approxumation. The idea is that we construct an equicontinuous sequence, so we extract a convergent subsequence, which is the solution.
Robin then proceeded to tell a story about a time when he visited Hershey Park in Hershey, PA and went on a rollercoaster.8 This somehow relates to the five-body problem where solutions escape to infinity in finite time. There is a Cantor set of initial conditions under which planets escape to infinity in finite time. Jared asked him what that would mean physically, and Robin started talking about shockwaves, which he claims to do for a living. He argues that it is more physically relevant to study integral equations than differential equations.
5.4.1 Differential Equations Continued
Theorem 72.
Let be the rectangle . Let be continuous on , so there exists such that on .
- Time of existence
-
Let . Then the solution exists and stays in for , i.e. for all .
- Uniqueness
-
If is Lipshitz in and uniformly continuous for , then the solution is unique for .
Proof.
- Time of existence
-
Define . Clearly , and also if and , then . Therefore is an interval. is closed because is continuous. ) Also, is relatively open in . If and , we have for . Therefore
Therefore there exists such that for . Therefore there exists a neighborhood of . Therefore .
- Uniqueness
-
Consider two solutions . Then we see that
where is the unoform Lipshitz constant. Therefore set . It satisfies the inequality for . This is a linear integral inequality. We will apply “Gronwall’s Lemma.”
Lemma 73 (Gronwall).
If and are continuous and defined on , with ans satisfy
then for we have .
Applying Gronwall here, for all . In particular, if , then for all . □
Note that we have proved continuous dependence on the initial conditions for short times. In particular, chaotic systems are predictable for a short time, after which they become unpredictable.
Proof of Gronwall.
Set . Then we know that for all . We differentiate to obtain . Also if , then . We see that , so . We can now take logs to obtain for all ,
or . Therefore we integrate to find that . Therefore . If , then we can’t do this. Instead replace with and let . □
Alternative existence proof for differential equations.
This proof uses Picard iteration. Remember the initial equation was . We integrate to obtain
This is an implicit nonlinear integral equation, but there is a natural iteration, which is Picard iteration.
Set . This is explicit and so convergence of the iteration to yields a solution. Then abstractly: Let be a complete metric space (think ). Then define the operator given by . Then solves the ODE if and only if it is a fixed point of . We look for an easy condition on a map which yields a fixed point.
Definition 74 (Strict Contraction).
A nonlinear map is a strict contraction if there exists such that for all , . Then we use the following theorem:
Theorem 75 (Contraction Mapping Theorem).
A strict contraction on a complete metric space has a unique fixed point.
We show that the map given above is a strict contraction. Then we see that
so if , this is a contraction. □
Proof of contraction mapping theorem.
For uniqueness, note that if there are two fixed points, then that contradicts the definition of a strict contraction.
For existence, just iterate. For any , set . We show that this sequence is Cauchy. After that, because is complete, the sequence converges, and its limit is a fixed point by continuity. To see that the sequence is Cauchy, this is simply because geometric series converge (this is left as an exercise because I don’t feel like typing). □
6 Banach Spaces
6.1 Lecture 10 (Feb 26)
We began the proof using Picard iteration last time, and Robin is leaving that as an exercise. We begin the discussion of Banach spaces, which are covered in chapters and of the textbook. Robin talked about proving estimates. We will add arithmetic structure to our metric space.
Recall the definition of a normed vector space :
Definition 76 (Normed Vector Space).
A normed vector space is a vector space satisfying:
- 1.
, ;
- 2.
;
- 3.
.
The extra structure here is linearity, which allows us to translate and scale, so the entire topology is determined by the unit ball.
Definition 77 (Banach Space).
A Banach space is a complete normed vector space.
Example 78.
The first example is with any -norm:
The power of is forced on us by the second axiom, and the triangle inequality is not at all obvious. Then for , define , which is the limit of the -norms.
We examine for different -norms, which has boundary . For we get the cross-polytope, the circle, and for we get the square. For we lose convexity of the unit ball, but for convexity holds, which is equivalent to the triangle inequality.
Note that the unit balls are nested for but inclusion of unit balls reverses inequalities in sizes of norms.
Example 79.
We have already seen that with the norm is Banach. However, note that is not closed with respect to the norm. The reason is that the norm cannot control derivatives.
To make Banach, we include the derivatives in the norm:
An alternative is to use . The two norms are equivalent.
Definition 80.
are equivalent as Banach spaces if the norms are equivalent (they bound each other above and below up to some constant). Clearly equivalent norms generate identical topologies.
Example 81.
Ww can extend norms on to infinite dimensions: Let be the set of sequences such that is finite. This is Banach with the -norm.
Example 82.
If , is Banach with norm the integral to the power.
Example 83.
If is a finite measure space, then we can define . Then this space with the obvious norm is a Banach space.
Note that the assumption that the space has finite measure is a convenience to ensure completeness. Otherwise, we must avoid issues of support “escaping to .”
Our goal is to show that is a Banach space. Before we do this, we introduce Sobolev spaces:
Definition 84 (Sobolev Space).
Define the Sobolev space to be:
Set the norm to the only possible norm.
Note that here we are cheating because we don’t know if the derivatives exist. We use a notion of weak derivative.
Theorem 85.
is a Banach space.
Proof.
Non-negativity (up to sets of measure ) and scaling are obvious. We need to prove the triangle inequality. Recall Jensen’s inequality:
Theorem 86 (Jensen’s Inequality).
Define the average to be
Then if is convex, .
Proof of Jensen was on the final exam last semester, so it is omitted.
Now recall Cauchy-Schwarz:
Theorem 87 (Cauchy-Schwarz).
If are square-integrable, then
6.2 Lecture 11 (Feb 28)
Hölder’s inequality generalizes this to :
Theorem 88 (Hölder’s Inequality).
Let and be the Hölder conjugate . Then if , then and
Proof.
If or this is obvious. We can assume that everywhere. The idea is to use Jensen. We will apply Jensen with , which is convex. Set , so . Note that by construction, is the Radon-Nikodym derivative . Therefore, by Jensen, we have
Taking roots and simplifying, we obtain
if we set . □
Corollary 89.
Assume . Then .
Proof.
Set . Thus . Therefore if , then . Also, since , for all , and in particular . Now we can write
Thus . □
Now we can prove Minkowski’s inequality:
Theorem 90 (Minkowski’s inequality).
If , then and .
Proof.
Firstly, note that , so .
This means that , so . Therefore
Therefore,
which gives the triangle inequality. □
This implies that is Banach. To prove is Banach, use the completeness of . If is Cauchy in , then is Cauchy in , so we take the power of the limit. □
An alternative approach is to use Young’s inequality , which is not Young’s convolution inequality.9
We return to an abstract Banach space . Unless explicitly stated otherwise, assume that is separable. Then we say has a Schauder basis if:
Definition 91 (Schauder basis).
has a Schauder basis if it has a countable set such that for all , we can approximate by .
Example 92.
The standard example is Fourier series, which can be uniquely approximated by .
Consider a linear operator .
Definition 93 (Bounded Linear Operator).
is bounded if there exists such that for all , we have .
The set of bounded linear operators is a normed vector space with the norm given by
. This is equivalent to .
Write to mean the space of bounded operators. If are Banach, so is .
Theorem 94.
is bounded if and only if it is continuous at .
Proof.
It is easy to see that boundedness implies continuity. To prove the converse, suppose not. if is unbounded, then for all , there exists such that . Take to have length . Then , so is not continuous at . □
We actually ended on time today, and Robin made a remark about me putting it in the notes.
6.3 Lecture 12 (Mar 5)
Robin apologizes for not assigning enough homework. He will assign homework this week, which will be due after break. There will be no midterm, unlike last semester.10
Let be a linear map of normed vector spaces. We see that the kernel and image are clearly subspace, and if is bounded, then is closed (by continuity). is the domain of . Note that changing the domain changes the operator. In a finite dimensional space, all linear operators are bounded (the closed unit ball is compact).
Lemma 95.
The closed unit ball is compact if and only if .
Proof.
In finite dimensions, the closed unit ball is closed and bounded, so it is compact.
In infinite dimensions, this follows from Riesz’s Lemma. We will find a sequence in which has no convergent subsequence. Take . For each , apply Riesz’s Lemma to . is not dense, so there exists such that and . Then no subsequence of can converge. □
Lemma 96 (Riesz).
If is a subspace, then if is not dense, then for all , there exists such that and .
Proof.
First choose . Let . We will scale and translate to get a unit vector. Choose and choose such that . Now set . Now we see that
Now choose small enough such that . □
Example 97.
In with any norm, the derivative is unbounded (take for example). Similarly, if we consider as a Schauder basis on , .
Example 98.
On , there are two common integral operators:
- Fredholm
-
, where is some kernel. is bounded in the norm if is finite. We see that
- Volterra
-
. Again . Note that the range of is not closed because is smoothing .
Example 99.
Let with the sup norm. The right shift and left shift are clearly bounded. is not onto and the image of is closed. Also, . An example of this is integrals and derivatives of power series with basis .
Theorem 100 (BLT).
If is a normed vector space, Banach, and a dense linear subspace of , then any bounded operator has a unique extension such that for all and .
Proof.
Clearly we define to be the limit where . Clearly this is well-defined (otherwise boundedness is contradicted). Also note that is clearly bounded by boundedness of and that for . By continuity, this is linear.
To prove uniqueness, let extend . Then by continuity, they must be equal ( is dense). □
Remark 101.
The above theorem allows us to pass to the limit from, for example, .
Definition 102 (Linear Functional).
A linear functional on is a bounded linear map .
Because is complete, the set of such functionals is complete whether or not is complete. We call this space , the dual space of .
6.4 Lecture 13 (Mar 7)
Robin began class questioning the point of shoveling snow. He says that it is necessary so you can drive your car away from your house.
Let be the space of bounded linear maps . A natural question is: when can we invert ? Robin drew a diagram and mentioned something about us not taking his linear algebra class. In , we have the transpose (note that this is not possible unless we have an inner product).
Theorem 103 (Rank-Nullity).
For finite dimensional vector spaces , we have , .
The restriction of is invertible as a map . In finite dimensions, a necessary algebraic condition for solvability of is . This is not enough. In our context, we want a bounded (continuous) inverse.
Theorem 104.
A bounded map of Banach spaces has a bounded inverse if and only if and is closed in .11
Proof.
One of the directions is easy. To prove the converse, we use the open mapping theorem.
Assuming the open mapping theorem, we note that open implies continuous. □
Theorem 105 (Open mapping theorem).
A bounded surjective map is open.
Sketch of Proof.
We show that is open. Since is surjective, write . Then apply the Baire Category Theorem. Therefore there exists such that has an interior. Eventually, we prove that there exists such that . □
Corollary 106.
If is a vector space with norms and such that and if is complete with respect to both norms, then the norms are equivalent.
Note that has more structure, namely composition. In particular, is a Banach algebra. This allows us to consider nonlinear functions of .
Example 107.
Some examples are exponentials and the Neumann series. To define the exponential, if , then we can define for any by a power series. It is easy to see that the series is absolutely convergent. Some properties of the exponential are that and is .
The Neumann weries is defined as the inverse of the operator . This is given by the usual geometric series.
In general, in the Bamach algebra , we can extend for any given function with convergent power series.
Definition 108.
An operator is compact if for any bounded set , the image is precompact in .
Sometimes, is too strong of a norm for our purposes. For , converges uniformly if . We will try to weaken this to broaden the class of convergent sequences. The first obvious weakening is pointwise convergence.
Example 109.
Consider . Then let be projection onto the first components. We see that for all , so is not Cauchy. However, pointwise.
Definition 110 (Weak convergence).
We weaken this notion further by declaring weakly, written if for each , in . Thus we use weak convergence in to define weak convergence in .
Definition 111 (Weak convergence in ).
weakly in if for all , in .12
6.5 Lecture 14 (Mar 19)
We continue our discussion of dual spaces. Let be a normed vector space and be Banach. Then , the space of bounded linear maps inherits completemenss from . Two of the most important are and . Denote the algebraic dual space of by and the topological dual by . Note that with the inclusion an equality when is finite-dimensional and a strict inclusion when is infinite-dimensional.
Lemma 112.
and are isomorphic when is finite dimensional.
Proof.
Choose a basis of . Then any satisfies , so every is determined by its values on the basis. Choose a dual basis given by . Clearly the are lienarly independent and any can be written as
by the expression for we computed above. The isomorphism is obvious now that they have the same dimension. □
Remark 113.
Note that this isomorphism is basis-dependent.
Example 114.
If is infinite dimensional, start with a countable independent set . Extend to a Hamel basis. Now given , write and note the usual way. Now form . Then but is unbounded.
Example 115 (Dual of ).
The answer is easy to find but hard to prove. Use Hölder’s inequality. If are conjugate and if , then . Thus we expect that for .
Sketch.
The first step is to show the equality . Take . Scale down with the appropriate powers and checking the desired equality is left as an exercise.
The next step is to show that . Start with and assume has finite measure. Note that for all . For measurable , define . Then . Also , where .
We show that is a signed measure and that . It is clear that is finitely additive. Similarly, if a disjoint union, write . Then . Therefore
From above, , so if , . Also since , as . Thus is countably additive.
Now we use Radon-Nikodym to obtain that there exists such that for all measurable, . Thus . Now extends to simple functions and then to measurable (and integrable) functions by taking limits. Therefore if for . Then if ,
Together with the first staement, this implies and that the norms are equal. □
Remark 116.
The assumption is essential, and in fact for of finite measure, any finite measure on with finite variation yields a bounded linear functional: . For general , we require Radon measures.
So far, we have and is complete. Note that is self-dual and we know that is reflexive for : . It is easy to show in general that : . Clearly this is bounded because , so is a bounded linear operator on .
Linear functionals are very useful geometrically and analytically. We use them to define weak convergence and weak derivatives. A question is: can separate a convex set from its exterior? if is convex and , is there some such that for and ?
6.6 Lecture 15 (Mar 21)
Robin apologizes for not returning our homework. He’s so organized.
We work in a vector space. Let be bounded convex and . We want to find a linear functional separating from such that for and . Note defines a “hyperplane.” Thus there is no loss of generality in taking . We will build a “Minkowski gauge function” by scaling around .
We will define . Then if and if . Also for all and for all and .
A function satisfying the last two conditions is called sublinear. The idea of Hahn-Banach is to extend a linear functional defined on the ray through to a linear functinal on all of , still dominated by .
Theorem 117 (Hahn-Banach).
Let be a vector space and sublinear. Then let be a subspace and is a linear functional on such that for all . Then extends to a linear functional on all of , such that for all and for all .
Proof.
First we show that we can extend by one dimension, then we close the process using the axiom of choice. Take . Let . Then is defined on , so we need . Note for . It is not enough to set and extend by linearity. Choose two points and . Consider
Now divide by and separate:
In conclusion, there exists such that
Now set and extend by linearity. Undoing this, we get for all .
For the second step, use Zorn’s lemma. Consider all extensions . Partially order them in the obvious way (subspace and restriction agrees). Now consider a chain . Then take . This is an upper bound for the chain. Therefore the set of extensions has a maximal element. Call it . Clearly this must be all of (otherwise apply step ). □
Remark 118.
We can prove Hahn-Banach without requiring the axiom of choice.
Corollary 119.
Let be a normed vector space and , and suppose . Then there exists extending such that .
Proof.
Set and apply Hahn-Banach □
Corollary 120.
For , there exists such that .
Proof.
Set . Apply the previous corollary. □
Corollary 121.
Let be a subspace. Let with . Then there exists a linear functional satisfying . , and for all .
Theorem 122.
If is separable, then is separable.
Proof.
If is dense in , for each , choose such that and . Consider
We show .
Suppose not. Then there exists with . By Corollary 121, there exists such that for all . Because , there exists a subsequence . Then
. This implies that , which is a contradiction. □
Remark 123.
The converse of the above is not necessarily true. For example is separable but is not separable.
7 Hilbert Spaces
7.1 Lecture 16 (Mar 26)
We will add more structure to our vector spaces. We revert to complex vector spaces. These are not conceptually harder, just computationally harder. Our inspiration comes from Pythagoras. The Euclidean distance comes from the dot product in .
Definition 124 (Inner Product).
An inner product in a complex vector space is a “sesquilinear form,” i.e. a map such that:
- 1.
where ;
- 2.
(Hermitian Symmetric);
- 3.
. In particular, ;
- 4.
if and only if .
We observe that the inner product is conjugate linear in the first variable and that sesquilinearity is needed to ensure non-negativity. We refer to equipped with as an inner product space.
Definition 125 (Hilbert Space).
A Hilbert space is a complete inner product space.
Example 126.
Consider complex matrices equipped with the inner product . This is called the “Hilbert-Schmidt norm.”
Example 127.
Consider with functions to . Then set . This is an inner product space, but is not Hilbert in general. We simply take the completion.
Example 128.
Now work with complex valued, -times differentiable . Set
and set to be the completion of .
Lemma 129.
An inner product induces a norm via .
Proof.
Linearity is obvious () for all and nonnegativity is also trivial. The triangle inequality follows from Cauchy-Schwarz. □
Theorem 130 (Cauchy-Schwarz).
For all , we have with inequality if and only if are linearly dependent.
Proof.
Set . Then
This implies the desired result. □
Theorem 131 (Parallelogram Law).
In any inner product space, .
Remark 132.
In fact, a norm on a vector space is induced from an inner product if and only if the parallelogram law holds.
Corollary 133.
is not Hilbert unless .
The law of cosines holds in any inner product space. More importantly, are orthogonal if and only if .
Theorem 134.
Let be a subset of a Hilbert space. Define . Then is a closed linear subspace of .
Proof.
It is clear that this is a subspace. Closedness is a consequence of continuity of the inner product. □
A partial converse is:
Theorem 135.
If is a closed linear subspace of . Then:
- 1.
For all , there exists a unique such that ;
- 2.
is the unique element such that .
Proof.
Recall that in finite dimensions, we simply use a sequence of projections. Set . There exists a sequence such that . We will show that the limit exists and that this element is unique. □
7.2 Lecture 17 (Mar 28)
We work in a Hilbert space with a closed subspace. We conclude the proof of Theorem 135 from last time.
Proof of Theorem 135, continued.
We set . There exists a sequence such that . We will show that the limit exists and that this element is unique. Consider the triangle with vertices . We will use the parallelogram law:
Therefore the sequence is cauchy and converges, so exists.
To show uniqueness, suppose there are two such points . Then .
We show that . For all , we have that . Let . Then setting , we get that . Dividing by , we have for all nonzero , which is only possibleif .
To show uniqueness, suppose there is another such . We know that . Using the parallelogram law, we know because . However, we see and thus . □
Remark 136.
We will reserve the direct sum for an orthogonal direct sum because if is closed with Banach, there may not be a closed space such that , . In a Hilbert space, for any set , we can write .
Definition 137 (Orthogonal Subset).
is an orthogonal subset if for all .
Definition 138 (Orthonormal Subset).
is orthonormal if is orthogonal and for all .
Example 139.
Consider -periodic functions, or functions on . Then for are orthogonal in . This forms a Fourier basis.
Example 140.
We can also consider quasi-periodic functions. is quasi-periodic if it is the sum of finitely many periodic functions. Let be the space of periodic functions of the form . We can define an inner product on by . We see is not complete. The completion of consists of almost periodic functions . Note that are orthogonal for . Thus we have an uncountable orthogonal set and thus is not separable.
In general we have an uncountable basis. Abstractly, we must make sense of where could be uncountable.
Definition 141 (Unconditional Convergence).
In a Banach space, we’ll say the unordered sum converges if the partial sums converge where is finite: there exists finite such that for any
We define absolute convergence like we do for .
Lemma 142.
Absolute convergence implies unconditional convergence. Also, for all there exists finite such that and set , which is countable. We see that for all .
Definition 143.
An unordered sum is Cauchy if for all there exists such that for all finite such that .
Lemma 144.
In a Banach space, an unordered sum converges if and only if it is Cauchy.
Proof.
Suppose the sum converges. Then there exists such that for any finite . IF is finite, set . Then .
Now suppose the sum is Cauchy. Choose increasing such that whenever is finite. Then , so is Cauchy. Thus it converges to some . Now given , choose such that . Then set , so if is finite, then □
7.3 Lecture 18 (Apr 02)
Recall the definition of unconditional convergence. Our goal is Bessel.
Lemma 145.
If is an orthogonal set in , then converges unconditionally if and only if , and in this case .
Proof.
For any finite set , . Therefore unconditional convergence is equivalent to Cauchy inplies that is Cauchy if and only if is Cauchy in . □
Corollary 146.
converges unconditionally in if and onlf if for all but countable many .
Theorem 147 (Bessel).
If is an orthonormal set in , then
- (a)
for all .
- (b)
converges unconditionally for all l
- (c)
.
Proof.
We must understand the sum in (b).
Therefore, for any finite , we have
This implies that converges if and only if converges.
If , then
□
We will write for the closed linear span of for an orthonormal set .
Theorem 148.
.
Theorem 149.
Let be an orthonormal subset of . Then the following are equivalent:
- (a)
for all only if ;
- (b)
for all ;
- (c)
for all ;
- (d)
;
- (e)
is a maximal orthonormal set in .
We call such a an orthonormal basis for .
Proof.
- (a) implies (b)
-
Note that . Bessel implies that .
- (b) implies (c)
-
In the proof of Bessel.
- (c) implies (d)
-
(c) states that . Thus , so .
- (d) implies (e)
-
If there exists such that , then since , for some such that . Since we must have and thus .
- (e) implies (a)
-
This is obvious.
□
We conclude that if we have an orthonormal basis, any can be componentwise.
Theorem 150.
Any Hilbert space has an orthonormal basis.
Proof.
Use Zorn’s lemma. Consider the collection of orthonormal subsets ordered by inclusion. Given a chain, the union of all sets in the chain is an upper bound. Thus the collection has a maximal element. By Theorem 149, it is an orthonormal basis. □
7.3.1 Adjoints
We return to Banach spaces. Let be a bounded linear map. We want to define a transpose. Recall the dual . Thus is given by evaluation. Then given a map we define an adjoint by . We see that is linear and bounded and satisfies .
Recall rank-nullity. Then analogous statements hold for . We introduce some notation: Given a subset , the annihilator is the subspace of such that if and onlf if for all . Then the range of is closed in because is bounded. We want that . Thus .
If , then for some . Under the identification , we have .
7.4 Lecture 19 (Apr 04)
Last time we defined the adjoint on a Banach space. Robin is going to do a lot of handwaving as he talks about integration by parts. To motivate this discussion, consider (our spaces have not yet been properly defined). We consider the space of compactly supported functions on . These are dense in all interesting spaces and vanish at the boundary.
Now define norms which include derivatives: in the norm . Then and our map is , which is differentiation. We see that differentiation is linear and that . We attempt to find the adjoint using . Then the adjoint of satisfies , which must be
We conclude that should be , whatever this means. Note the domain of should be bigger than .
Remark 151.
Our couplings used to define adjoings are in some sense based on the inner product.
In a Hilbert space , we have a natural pairing, which is given by the inner product. We would lise to define the adjoint of by .13 This is consistent because any Hilbert space is self-dual.
Theorem 152 (Riesz Representation Theorem).
Let be a Hilbert space. Then for all there exists a unique such that .
We develop some machinery to prove Riesz. Let be a vector space. Then is a projection is . We can enforce an orthogonal projection if . In a Hilbert space, . For a proof, consider the inner product of with itself. Define the reflection by . Then are projections onto even and odd parts of , respectively.
Proof of Riesz Representation.
If then we take . Thus assume . Then there exists such that and . Define by . It is clear that is a projection. Also note that if and only if . Thus , which is an orthogonal direct sum.
Writing , we see that . Thus we want such that . We see that . Then for all , we can calculate that .
To show uniqueness, note that for all if and only if . □
7.5 Lecture 20 (Apr 09)
Last time we proved that a Hilbert space is self-dual. Define by the mapping given by the Riesz Representation Theorem. Adjoints in Hilbert spaces can now be defined using the inner product and we can define the adjoint by . We want to generalize the rank-nullity theorem. Because , then implies .
Theorem 153.
Let be linear. Then and .
Proof.
If , then . If , then for all . Thus . However we know that the closure of contains . This gives the first part.
For the second part, apply the first part to . □
This theorem provides a necessary solvability condition for the abstract equation , which is that is orthogonal to the kernel of . One context for this is partial differential equations. We take a PDS and look for perturbations of some known solution: . Then we collect powers of , which typically leads to separate PDEs for . The goal is to approximate the solution by . We need to have some bounded solution, which is where we write the solvability condition.
The second application is the relation between uniqueness and existence. Suppose that the adjoint problem has a unique solution. Then Thus by the theorem, if has closed range, then has a solution for any .
Lemma 154.
If has a unique solution and if for some , then has a solution for all .
Definition 155.
has the Fredholm alternative if either:
- 1.
implies , and both and have unique solutions;
- 2.
and have nontrivial finite-dimensional solution spaces of the same dimension, and has a solution iff for all such that , and the same holds for .
Definition 156.
is a Fredholm operator if and
- 1.
is closed;
- 2.
Both and are finite dimensional.
Define the index of as .
Proposition 157.
If is Fredholm and is compact, then is Fredholm.
We will now discuss self-adjoint operators.
Definition 158.
is self-adjoint if .
Self-adjoint operators satisfy the spectral theorem.
Lemma 159.
If is self-adjoint, then:
- 1.
;
- 2.
, and if is self-adjoint, then ;
Theorem 160 (Spectral Theorem, compact case).
Let be compact and self-adjoint on . Then there is an orthonormal basis of eigenvectors: The non-zero eigenvalues form a countable set of reals, and we can write where each is an orthogonal projection onto the finite-dimensional eigenspace corresponding to . The series converges in operator norm.
7.6 Lecture 21 (Apr 11)
We consider the spectral theorem again. Recall the definition of a Fredholm operator. Note that is invertible, and so is . If is self-adjoint, then the three conditions coincidem and the only requirement is that the kernel is finite-dimensional. In particular, if is invertible, it is Fredholm.
Recall the definitions of eigenvalue and eigenvector.
Proposition 161.
If is self-adjoint, then
- 1.
All eigenvalues are real.
- 2.
Eigenvectors corresponding to different eigenvalues are orthogonal.
Most of what we still will hold for normal operators, but we state them for self-adjoint operators.
Definition 162 (Normal Operator).
An operator is normal if it commutes with its adjoint.
Example 163.
Skew-adjoint operators () are normal.
Lemma 164.
Let be self-adjoint. Then .
Sketch of Proof.
Let be the supremum. Then follows easily. Then note that , where have norm . We can use the parallelogram law and choose by a factor of . □
Definition 165 (Resolvent).
Define the resolvent of as . Note that because is bounded, the inverse is also bounded. The spectrum of is .
There are two cases for :
- 1.
is Fredholm. Then there is a finite eigenspace and thus is an eigenvalue of finite multiplicity. The discrete spectrum is For each such , there are a finite number of eigenvalues that span a finite dimensional space , and is the projection onto that space.
- 2.
is not Fredholm. Define the essential spectrum . We know that
- (a)
;
- (b)
is closed in ;
- (c)
If is self-adjoint and compact, then .
- (a)
Theorem 166.
The following are equivalent:
- 1.
.
- 2.
Weyl’s Criterion: there exists a sequence in with such that , but has no convergent subsequence.
- 3.
, or there exist such that .14
Remark 167.
Robin believes that if is normal, then there is no essential spectrum.
8 Fourier Series
Consider periodic with period . We can regard . This becomes a Hilbert space with . The Fourier basis is the orthonormal basis where . To get a hilbert space, we set in the corresponding norm.
Proposition 168.
is a complete orthonormal basis for .
We conclude that any function can be written
where . This sum converges provided . Using Euler’s identity, we can write in the real Fourier sine/cosine series
We have and . Also . The sine and cosine series are particularly nice for even and odd functions. A property of plays very nicely with the derivative and is in fact an eigenfunction of the differentiation operator.
8.1 Lecture 22 (Apr 16)
We continue with our discussion of Fourier series. We will end lecture by solving a PDE. For simplicity, our period is . We want to determine when and in what sense a Fourier series converges.
First, we see that converges if , as in . Also, if , we get convergence.
Proposition 169.
is a complete orthonormal basis on .
The idea of the proof is to set and follow the Weierstrass approximation theorem. For details, see the text.
8.1.1 Distributions and Fourier Transform
We will begin discussing Fourier transforms, which are more general than Fourier series.
Remark 170.
Fourier transforms commute with convolution.
Theorem 171 (Carleson, 1966).
If , then pointwise almost everywhere, However, convergence fails at points of discontinuity of .15
Note that the power of Fourier series and the FFT is the speed with which this is calculated, which is . Wavelets are also efficient, but they are discontinuous, so they avoid Gibbs phenomena and can be used in image processing.
We know that , so . Differentiating the Fourier series just scales the modes, so differentiating times should lead to the Fourier series of -th derivatives. Thus is -times differentiable if and only if converges. This is one way to define the Sobolev space .
Definition 172 (Fourier Transform).
Let . Then the Fourier transform is
The inverse transform is given by
We think of being physical space and being frequency space.
Theorem 173 (Plancharel).
If , then and in .
Idea of Proof.
By our assumptions, the transform exists everywhere. Then we compute
If have compact support, then we can integrate to get . This all needs to be checked.
Taking this for granted, we get that the norm is . □
It follows that we can extend the Fourier transform to be defined on all of . To do this, we approximate be and take limits.
Proposition 174.
- 1.
.
- 2.
.
- 3.
.
- 4.
If , then for .16
8.2 Lecture 23 (Apr 18)
We will define the notion of a distribution, following the note from Smoller on the webpage.
Definition 175 (Distribution).
A distribution is a continuous linear functional on the closure of , which is the space of compactly supported smooth function.
Here we have avoided the mention of a topology because is not metrizable. We define the strong topology:
Definition 176 (Strong Topology).
We declare in the strong topology if for all nonzero nulti-indices.
We call the test dunctions. Then the disributions are continuous in the weak topology, where if for all .
Example 177.
- 1.
If , then .
- 2.
Evaluation at , given by .
- 3.
Any Radon measure gives .
- 4.
Evaluation of derivatives: For multiindex, .
We define the support of a distribution .
Definition 178.
We say vanishes on a set if for all whose support is a subset of . The support .
8.2.1 Weak Derivatives
These are defined by integration by parts. For example, if and is a multi-index, is the -weak derivatives if integration by parts holds on all smooth test functions . That is, is the -weak derivative of if for all test functions , .
We can define this for any distribution.
Definition 179 (Weak Derivative).
Let be a distribution and a multi-index. Then is defined by .
Note that we cannot multiply distributions, but we can multiply distributions by functions by .
Example 180.
If is continuous at , we can define . However, if is the Heaviside function, this does not make sense because , but , so we don’t know what to use.
An application of this is Burger’s equation: . This admits discontinuous solutions, so we can write it in the better form .17
Recall that convolutions are like products. Then we can define
We can rewrite this as . We can use this to define convolutions for distibutions.
8.3 Lecture 24 (Apr 23)
Recall the notion of a distribution. Today we will define the convolution of two distributions.
Definition 181 (Convolution of Distributions).
The convolution of two distributions is defined by . Here, we require at least one of has compact support.
Lemma 182.
Suppose is a distribution and is a compactly supported distribution.
- 1.
is a distribution.
- 2.
- 3.
If is a compactly supported distribution then
- 4.
.
- 5.
If and is a distribution, then regarding as a distribution, .
Remark 183.
Note that the delta function is the unit with respect to convolution.
Lemma 184.
Convolution commutes with differentiation.
Proof.
□
We will use the same tricks to define the Fourier transform of some distributions. We have too many distributions, so we extend our class of test functions.
Definition 185.
Schwartz space is the set of those such that for all multi-indices ,
Any such that is in .
Note the following properties of multi-indices.
- 1.
Taylor:
where and .
- 2.
Leibniz:
We consider the topology of . Set . This is not quite a norm but is a seminorm. Given a family of seminorms on , we define a topology on by is an open neighborhood of if . The separate points if implies for all . If the family separates points, the topology is Hausdorff. In this case is locally convex.
is also metrizable if is countable. We need to take account of all the seminorms:
Theorem 186.
In this metric, is complete.
Proof.
Consider a Cauchy sequence . Then . Also by definition of the , the sequence is Cauchy in . Thus the limit is continuous. Then using we get a Cauchy sequence for the derivatives. We need to prove that the limits coincide with the derivarives of the limit .
We do this by induction on . Suppose . Then there exists such that where . Now write
Taking the limit as , we see that
Dividing by and taking , we see that by continuity. This gives that is Cauchy and thus converges to . □
8.4 Lecture 25 (Apr 25)
We continue our discussion of Schwarz space. We will apply the Fourier transform to and see what happens. Recall that the Fourier transform morally decomposes a function into its pieces which look like with wavenumber . We know and . In fact, if , then the Fourier transform is continuous.
Not we introduce the translation operator . Then a simple calculation shows that . Assuming is , note that the Fourier transform of is . Taking the limit as , we conclude that .
Next we consider the Fourier transform of a convolution. This was on the homework last semester, so the calculation is omitted (just use Fubini):
Define . We will calculate
We know that . Then formally, . In order to justify this, we will need to consider tempered distributions. The tempered distributions are elements of , the dual of . Note that . Thus maps to itself. Assuming are nice enough, we can see that . Therefore, given , we define by .
We conclude the calculation with
The sine term is a kernal approaching as , which will be confirmed later.
8.4.1 Yet another differential equation
Consider a linear hyperbolic PDE . Formally, we apply the Fourier transform. First we solve the easier problem , where is some element of the algebra of differential operators. We can apply the Fourier transform here, find the Fourier transform formally, and then invert. Because , solves .
8.5 Lecture 26 (Apr 30)
We go back to the inversion formula.
Theorem 187.
If then .
Proof.
First restrict to . By Fubini, follows immediately. Then for we have
where . Therefore . We know that (proved using complex analysis). Thus . Then for large, we have as distributions. To justify this, we calculate:
Integrate by parts to get
because . Now we see that
The first integral yields again, and the second integral becomes as . □
We consider more examples. For example, if , this determines a distribution. However, if , can we still define a corresponding distribution? (As an example, take ). Generally, it is not really possible. We lose “balance” by using weighted artificially to one side of the singularity at . On the other hand, if is even, we can use the “principal value” distribution:
The principal value is defined by . We show that this is a distribution. Write
Now observe that , wo the integrand is bounded. Now write
Now consider the fimal example, which is the PDE from last time. Given a polynomial , its symbol is the polynomial . Observe that . Therefore a polynomial operator corresponds to a polynomial in Fourier space, and we can do this for more general functions. If is a function of , we can define an operator in physical space corresponding to , which extends polynomials to power series and allows so-called “Fractional Laplacians.”
For example, in , the Laplacian has . Thus we can define the operator by . This is a radially symmetric derivative. Similarly, the operator defined by is very useful for defining Sobolev spaces. The fractional Sobolev space can be defined by .