Advanced Topics in Functional Analysis

The Hahn–Banach Theorem and Spectral Theorem Introduction Two of the most powerful results in functional analysis are the Hahn–Banach Theorem and the Spectral Theorem. These theorems underpin much of modern analysis and are essential for understanding linear operators and functionals on infinite-dimensional spaces. The Hahn–Banach Theorem guarantees the existence of enough continuous linear functionals, while the Spectral Theorem provides a canonical representation of important classes of operators. The Hahn–Banach Theorem Understanding the Problem The fundamental question that motivates the Hahn–Banach Theorem is: given a linear functional defined only on a subspace, can we extend it to the entire vector space while preserving certain properties? Consider a subspace $U$ of a larger vector space $V$. Suppose we have a linear functional $f: U \to \mathbb{R}$ (or $\mathbb{C}$) that satisfies some upper bound condition. The Hahn–Banach Theorem tells us we can extend this to a functional $F: V \to \mathbb{R}$ (or $\mathbb{C}$) on the whole space while maintaining that bound. The Extension of Linear Functionals The Statement: Let $p$ be a sublinear function on a vector space $V$—that is, a function satisfying: $p(\alpha x) = \alpha p(x)$ for all scalars $\alpha \geq 0$ and $x \in V$ (positive homogeneity) $p(x + y) \leq p(x) + p(y)$ for all $x, y \in V$ (subadditivity) Suppose $f$ is a linear functional on a subspace $U \subset V$ such that $$f(x) \leq p(x) \text{ for all } x \in U$$ Then there exists a linear functional $F$ on the entire space $V$ such that: $F$ extends $f$, meaning $F(x) = f(x)$ for all $x \in U$ $F$ respects the bound: $F(x) \leq p(x)$ for all $x \in V$ Why This Matters: A Common Application A common special case occurs when $V$ is a normed vector space and we take $p(x) = \|x\|$. The theorem then guarantees that a bounded linear functional on a subspace can be extended to a bounded linear functional on the entire space, with the same operator norm. This is crucial because: Without such extensions, we might lose important functionals when moving from subspaces to larger spaces The theorem ensures that norms are preserved under extension It validates that our constructions on subspaces are compatible with the larger structure Key insight: The power of Hahn–Banach lies in its generality—it applies to any normed space, without requiring completeness or other special structure. Consequences for Dual Spaces One of the most important consequences is that every normed vector space $V$ has a large dual space $V^$ consisting of continuous linear functionals. Specifically, the Hahn–Banach Theorem guarantees: Separation Property: For any nonzero element $x \in V$, there exists a continuous linear functional $F \in V^$ such that $F(x) \neq 0$. (This shows the dual space is non-trivial.) Abundance of Functionals: The dual space is "rich enough" to distinguish between different elements of $V$. You cannot have two different elements of $V$ that are indistinguishable to all functionals in $V^$. Useful Geometric Structure: This allows us to study properties of elements in $V$ by examining how they behave under all continuous linear functionals—a duality that is fundamental to functional analysis. Without this theorem, we would have no guarantee that the dual space contains any non-zero functionals at all! The theorem transforms the dual space from something that might be trivial into a rich mathematical object. The Spectral Theorem Motivation: Diagonalization in Infinite Dimensions In finite-dimensional linear algebra, one of the most useful facts is that certain matrices (like symmetric matrices) can be diagonalized: they can be written as $A = PDP^{-1}$ where $D$ is diagonal. The diagonal form reveals the eigenvalues and simplifies computations. The Spectral Theorem is the infinite-dimensional generalization of this idea. However, because infinite-dimensional Hilbert spaces have more complex structure, diagonalization takes a different form: instead of a diagonal matrix, we use multiplication by a function. Bounded Self-Adjoint Operators An operator $T$ on a Hilbert space $H$ is self-adjoint if $T = T^$, where $T^$ is its adjoint. These are the infinite-dimensional analogs of symmetric matrices. The Spectral Theorem for Self-Adjoint Operators: For any bounded self-adjoint operator $T$ on a Hilbert space $H$, there exist: A measure space $(\Omega, \mu)$ (the "spectrum") A real-valued essentially bounded measurable function $f: \Omega \to \mathbb{R}$ A unitary operator $U: H \to L^2(\Omega, \mu)$ (an isometry that preserves the structure) such that $$T = U^ Mf U$$ where $Mf$ denotes the multiplication operator: $(Mf \phi)(\omega) = f(\omega) \phi(\omega)$ for functions $\phi \in L^2(\Omega, \mu)$. What This Means This representation says that every bounded self-adjoint operator is "unitarily equivalent" to a multiplication operator. Intuitively: The unitary operator $U$ provides a change of coordinates (an isometric isomorphism) In these new coordinates, $T$ simply multiplies by a function $f$ The function $f$ contains all the spectral information—its values represent "generalized eigenvalues" Since multiplication operators are diagonal in a sense (they act independently on each point in the domain), this is a generalized form of diagonalization Key advantage: Multiplication operators are much simpler to understand and analyze than general operators, so this canonical form is extremely useful for computations and proofs. Bounded Normal Operators An operator $T$ on a Hilbert space is normal if $TT^ = T^T$—that is, it commutes with its adjoint. Self-adjoint operators are a special case of normal operators. The Spectral Theorem for Normal Operators: An analogous statement holds for bounded normal operators. The only difference is that the representing function $f$ may be complex-valued rather than real-valued: $$T = U^ Mf U$$ where $f: \Omega \to \mathbb{C}$ is complex-valued and essentially bounded. The complex-valued function $f$ encodes information about how $T$ acts, including both the magnitude and phase effects. This generalization is important because many operators of interest (like certain differential operators) are normal without being self-adjoint.