To obtain a broad basis in functional analysis well beyond the introductory level, preparing for a specialization in fundamental analysis as well as developing the tools for advanced functional analytic applications in other disciplines. In order to cover a wide range of diverse topics, the lectures will often focus more on conceptual aspects rather than technical proofs, many of which will be sketched or omitted.
The course starts with a thorough introduction to topological vector spaces (TVS). In general TVS, a norm — and even a metric — is not available. However, many theorems well-known in the Banach space context still hold. Their proofs then take a much more topological flavour.
To smoothen the transition to non-metric vector spaces, we start by proving the Banach-Steinhaus Theorem (a.k.a. Uniform Boundedness Principle), first in the Banach space setting, then in the more general TVS context. This result is based on the Baire Category Theorem, other important consequences of which are also presented, such as the Open Mapping Theorem and the Closed Graph Theorem. We then study in more detail the important case of locally convex TVS, which are most common in applications, e.g. for the theory of distributions, which is briefly discussed.
The next topic is convexity and various versions of the Hahn-Banach Theorem. This is followed by the introduction of weak and weak* topologies, and related compactness results, such as the Banach-Alaoglu Theorem and the Eberlein-Smulyan Theorem.
The broader theme of the next part of the course is spectral analysis. We introduce the notions of spectrum, resolvent set and eigenvalues in an infinite-dimensional setting. As a first case, we study the spectral properties of compact operators on Banach spaces. Here, the situation is still very similar to the finite-dimensional setting. We prove the Spectral Theorem for compact operators, and discuss the approximation property of Banach spaces. Concrete applications to integral equations and Sturm-Liouville problems are considered in the homework problems.
With the aim of generalizing the Spectral Theorem to other classes of operators, we then have a look at Banach algebras and explain what the spectrum in this context means. We introduce the Dunford calculus, a symbolic calculus on Banach algebras, by means of Cauchy's Integral Theorem. We outline how this can be used in applications to differential equations.
In the next part of the course, we give a brief introduction into Gelfand theory on commutative Banach algebras. This includes the Gelfand transform as a generalization of the Fourier transform, and the Gelfand-Naimark Theorem. We then specialise the Gelfand theory to the study of bounded normal operators on a Hilbert space. In this context, we define a functional calculus and prove the Spectral Theorem for bounded normal operators.
The final part of the course deals with unbounded operators in Hilbert spaces, and in particular the notions of symmetric and selfadjoint operators. Selfadjointness for unbounded operators is a delicate matter, related to the domain of the operator. The theory of selfadjoint extensions of symmetric operators is presented, with concrete examples of differential operators discussed in the homework problems.
We then study the Spectral Theorem for unbounded selfadjoint operators. The proof we present, due to John von Neumann, is based on the Cayley transform. A notion of integration with respect to a spectral family is central to the Spectral Theorem. In the case of unbounded operators, it is deeply connected with the Lebesgue-Stieltjes integral. As an application, we conclude the course with Stone's theorem, which characterizes one-parameter groups of unitary operators. If time permits, we also briefly discuss applications to quantum mechanics.
Basic knowledge of bounded linear operators in Banach and Hilbert spaces, of general topology and metric spaces. Students should be familiar with notions such as Cauchy sequences in normed vector spaces, operator norm, dual space, inner product and Cauchy-Schwarz inequality, orthogonal projections in Hilbert spaces, orthonormal basis and Fourier coefficients, adjoint operator, selfadjoint, unitary and normal operators. This course is not suitable as a first acquaintance with functional analysis. Measure and integration theory is not a formal prerequisite, an intuitive knowledge of it will be enough in the beginning of the course. Later on, we will however assume that all participants are familiar with measure and integration theory at a workable level. It is essential that students who haven't yet had a course in measure and integration theory follow one in parallel to our course.
Rules about Homework / Exam
The global grade will be based on 30% homework assignment and 70% final written exam. In order to pass the course, the grade in the exam must be a 5 or higher.
There will be 4 substantial homework assignments during the semester.
The written exam will be a 3-hour exam, with no material allowed.