Physlib

PhyslibAlpha.ProbabilisticTheory.HilbertSpace.Unbounded.SelfAdjointSpectralTheorem

The spectral theorem, stated

i. Overview

An operator is essentially self-adjoint when its closure is self-adjoint; a self-adjoint extension is then unique. A spectral measure `μ` on `ℝ` resolves a self-adjoint operator `T` when `⟪y, T x⟫ = ∫ λ d⟪y, μ x⟫` for `x` in the domain of `T`, and it resolves `T` with its domain when in addition the domain of `T` is the set of vectors with `∫ λ² dμₓ < ∞`.

ii. Key results

- `SelfAdjointClosureData` : an operator with self-adjoint closure. - `IsWeakSpectralResolution` : `⟪y, T x⟫ = ∫ λ d⟪y, μ x⟫` on the domain of `T`. - `spectralSquareMomentDomain` : the vectors with finite second moment. - `DomainAwareSelfAdjointSpectralTheorem` : a spectral measure resolving a self-adjoint operator together with its domain.

iii. Table of contents

- A. Essential self-adjointness and closure - A.1. Spectral resolutions with domain

A. Essential self-adjointness and closure

An operator is essentially self-adjoint when its closure is self-adjoint.

A.1. Spectral resolutions with domain

A spectral resolution only describes the operator on its domain. Equality of unbounded operators also needs the domain to be the set of vectors of finite second moment.

21 declarations

theorem

The closure Tˉ\bar{T} of an essentially self-adjoint operator is self-adjoint

Let TT be an essentially self-adjoint operator. Then its closure, denoted by Tˉ\bar{T}, is a self-adjoint operator.

theorem

The closure Tˉ\bar{T} of an essentially self-adjoint operator is closed

Let TT be a linear operator such that its closure Tˉ\bar{T} is self-adjoint (i.e., TT is essentially self-adjoint). Then the closure Tˉ\bar{T} is a closed operator.

theorem

T≤TˉT \leq \bar{T} for essentially self-adjoint operators

Let TT be a linear operator such that its closure Tˉ\bar{T} is self-adjoint. Then T≤TˉT \le \bar{T}, which means that the domain of TT is contained in the domain of its closure Tˉ\bar{T}, and the two operators coincide on the domain of TT.

theorem

The unique self-adjoint extension of an essentially self-adjoint operator TT is its closure Tˉ\bar{T}

Let HH be a complex Hilbert space and TT be a linear operator on HH such that its closure Tˉ\bar{T} is self-adjoint (i.e., TT is essentially self-adjoint). If SS is any self-adjoint extension of TT (denoted T≤ST \leq S, meaning dom(T)⊆dom(S)\text{dom}(T) \subseteq \text{dom}(S) and SS restricted to dom(T)\text{dom}(T) equals TT), then S=TˉS = \bar{T}.

theorem

Zero Defect Numbers at ±i\pm i imply Essential Self-Adjointness

Let HH be a complex Hilbert space and TT be a densely defined symmetric linear operator on HH. If the defect numbers of TT at ii and −i-i are both zero—that is, dim⁡(Ran(T−iI)⊥)=0\dim(\text{Ran}(T - iI)^\perp) = 0 and dim⁡(Ran(T+iI)⊥)=0\dim(\text{Ran}(T + iI)^\perp) = 0—then TT is essentially self-adjoint, meaning its closure Tˉ\bar{T} is a self-adjoint operator. This result represents the von Neumann defect-number criterion for essential self-adjointness.

theorem

A defect index certificate for TT implies self-adjoint closure data for TT

Let HH be a complex Hilbert space and TT be a linear operator on HH. If there exists a defect index certificate CC for TT, then TT possesses self-adjoint closure data, which implies that the closure of TT is a self-adjoint operator.

definition

Weak spectral resolution ⟨y,Tx⟩=∫λ d⟨y,μx⟩\langle y, Tx \rangle = \int \lambda \, d\langle y, \mu x \rangle

For a linear operator TT on a complex Hilbert space HH and a spectral measure μ\mu on R\mathbb{R}, this property states that μ\mu is a weak spectral resolution of TT. Specifically, for every xx in the domain of TT, the identity function f(λ)=λf(\lambda) = \lambda must be integrable with respect to the complex measure μx,y\mu_{x,y} (defined as μx,y(A)=⟨y,μ(A)x⟩\mu_{x,y}(A) = \langle y, \mu(A)x \rangle) for all y∈Hy \in H, and the following relation must hold: ⟨y,Tx⟩=∫Rλ d⟨y,μ(λ)x⟩\langle y, Tx \rangle = \int_{\mathbb{R}} \lambda \, d\langle y, \mu(\lambda)x \rangle

definition

Domain of finite second moment ∫λ2 dμx<∞\int \lambda^2 \, d\mu_x < \infty

Given a spectral measure μ\mu on R\mathbb{R} acting on a Hilbert space HH, the set consists of all vectors x∈Hx \in H for which the second moment with respect to the spectral measure is finite. Specifically, a vector xx is in this set if the function λ↦λ2\lambda \mapsto \lambda^2 is integrable with respect to the scalar measure μx\mu_x defined by μx(A)=⟨x,μ(A)x⟩\mu_x(A) = \langle x, \mu(A)x \rangle. In integral notation, this is the set of x∈Hx \in H such that ∫Rλ2 d⟨x,μ(λ)x⟩<∞.\int_{\mathbb{R}} \lambda^2 \, d\langle x, \mu(\lambda)x \rangle < \infty. This set defines the natural domain of the self-adjoint operator T=∫λ dμ(λ)T = \int \lambda \, d\mu(\lambda) associated with the spectral measure.

theorem

x∈spectralSquareMomentDomain μ  ⟺  ∫λ2 dμx<∞x \in \text{spectralSquareMomentDomain } \mu \iff \int \lambda^2 \, d\mu_x < \infty

Let HH be a Hilbert space and μ\mu be a spectral measure on R\mathbb{R} acting on HH. For any vector x∈Hx \in H, xx belongs to the domain of finite second moment of μ\mu (denoted as `spectralSquareMomentDomain μ`) if and only if the function λ↦λ2\lambda \mapsto \lambda^2 is integrable with respect to the diagonal measure μx\mu_x (defined as μx(A)=⟨x,μ(A)x⟩\mu_x(A) = \langle x, \mu(A)x \rangle). In other words, xx is in the domain if and only if ∫Rλ2 d⟨x,μ(λ)x⟩<∞.\int_{\mathbb{R}} \lambda^2 \, d\langle x, \mu(\lambda)x \rangle < \infty.

definition

Spectral measure μ\mu is supported on [−C,C][-C, C]

Given a spectral measure μ\mu on R\mathbb{R} acting on a Hilbert space HH and a real number CC, this property states that the support of the spectral measure is contained within the closed interval [−C,C][-C, C]. Specifically, for every measurable set S⊆RS \subseteq \mathbb{R}, if SS is disjoint from [−C,C][-C, C], then the spectral measure of SS is zero, i.e., μ(S)=0\mu(S) = 0.

theorem

Bounded Spectral Support Implies the Domain of Finite Second Moment is the Entire Space HH

Let HH be a Hilbert space and μ\mu be a spectral measure on R\mathbb{R} acting on HH. If the spectral measure μ\mu has bounded support contained within the interval [−C,C][-C, C] for some C∈RC \in \mathbb{R}, then its domain of finite second moment (the set of vectors x∈Hx \in H such that ∫Rλ2 d⟨x,μ(λ)x⟩<∞\int_{\mathbb{R}} \lambda^2 \, d\langle x, \mu(\lambda)x \rangle < \infty) is the entire Hilbert space HH.

theorem

Bounded Spectral Support and Full Domain imply Domain-Aware Spectral Resolution

Let HH be a Hilbert space and TT be a linear operator on HH with domain D(T)\mathcal{D}(T). Suppose μ\mu is a spectral measure on R\mathbb{R} that resolves TT, meaning that for all x∈D(T)x \in \mathcal{D}(T) and y∈Hy \in H, the identity ⟨y,Tx⟩=∫Rλ d⟨y,μ(λ)x⟩\langle y, Tx \rangle = \int_{\mathbb{R}} \lambda \, d\langle y, \mu(\lambda)x \rangle holds. If the domain of TT is the entire Hilbert space (D(T)=H\mathcal{D}(T) = H) and the support of μ\mu is bounded (contained in some interval [−C,C][-C, C] for C∈RC \in \mathbb{R}), then μ\mu provides a domain-aware spectral resolution for TT. This implies that the domain of TT is exactly the set of vectors with finite second moment, i.e., D(T)={x∈H∣∫Rλ2 d⟨x,μ(λ)x⟩<∞}\mathcal{D}(T) = \{x \in H \mid \int_{\mathbb{R}} \lambda^2 \, d\langle x, \mu(\lambda)x \rangle < \infty \}.

theorem

Domain-aware spectral resolution implies TT is self-adjoint

Let HH be a Hilbert space and TT be an operator acting on HH. If μS\mu_S is a spectral measure that provides a domain-aware spectral resolution for TT (meaning TT is represented by the integral ∫λ dμS\int \lambda \, d\mu_S and its domain coincides with the set of vectors of finite second moment), then TT is a self-adjoint operator.

theorem

Weak Spectral Reconstruction from Domain-Aware Spectral Theorem

If TT is a linear operator on a complex Hilbert space HH and μ\mu is a spectral measure on R\mathbb{R} that provides a domain-aware spectral resolution for TT, then μ\mu is a weak spectral resolution of TT. Specifically, for every vector xx in the domain of TT and every vector y∈Hy \in H, the identity function f(λ)=λf(\lambda) = \lambda is integrable with respect to the complex measure μx,y\mu_{x,y} (defined by μx,y(A)=⟨y,μ(A)x⟩\mu_{x,y}(A) = \langle y, \mu(A)x \rangle), and the following identity holds: ⟨y,Tx⟩=∫Rλ d⟨y,μ(λ)x⟩\langle y, Tx \rangle = \int_{\mathbb{R}} \lambda \, d\langle y, \mu(\lambda)x \rangle

theorem

x∈dom(T)x \in \text{dom}(T) if and only if ∫λ2 dμx<∞\int \lambda^2 \, d\mu_x < \infty

Let HH be a Hilbert space, TT be an operator on HH, and μ\mu be a spectral measure on R\mathbb{R} that resolves TT. For a spectral resolution that accounts for the operator's domain (represented by `DomainAwareSelfAdjointSpectralTheorem`), a vector x∈Hx \in H belongs to the domain of TT if and only if the second spectral moment of xx with respect to μ\mu is finite. That is, x∈dom(T)  ⟺  ∫Rλ2 d⟨x,μ(λ)x⟩<∞.x \in \text{dom}(T) \iff \int_{\mathbb{R}} \lambda^2 \, d\langle x, \mu(\lambda)x \rangle < \infty.

definition

Strongly continuous unitary group U(t)=eitTU(t) = e^{itT} from the spectral theorem

Given a domain-aware spectral resolution for a self-adjoint operator TT with associated spectral measure μ\mu on a Hilbert space HH, this definition constructs the strongly continuous one-parameter unitary group U:R→B(H)U: \mathbb{R} \to \mathcal{B}(H). For each t∈Rt \in \mathbb{R}, the operator U(t)U(t) is defined via the spectral integral: U(t)=∫Reitλ dμ(λ). U(t) = \int_{\mathbb{R}} e^{it\lambda} \, d\mu(\lambda). This construction provides the unitary group U(t)=eitTU(t) = e^{itT} associated with the operator TT as described by Stone's theorem, satisfying the group properties U(0)=IU(0) = I and U(t+s)=U(t)U(s)U(t+s) = U(t)U(s).

theorem

The unitary group U(t)=eitTU(t) = e^{itT} satisfies U(0)=IU(0) = I

Let TT be a self-adjoint operator on a Hilbert space HH with an associated spectral measure μ\mu provided by a domain-aware spectral resolution. Let U(t)=∫Reitλ dμ(λ)U(t) = \int_{\mathbb{R}} e^{it\lambda} \, d\mu(\lambda) be the strongly continuous one-parameter unitary group associated with TT (often denoted eitTe^{itT}). Then, at t=0t = 0, the operator is the identity operator, i.e., U(0)=IU(0) = I.

theorem

Group property of the unitary group eitTe^{itT}: U(t+s)=U(t)U(s)U(t+s) = U(t)U(s)

Given a domain-aware spectral resolution for a self-adjoint operator TT with associated spectral measure μ\mu on a Hilbert space HH, let U(t)=eitTU(t) = e^{itT} be the strongly continuous one-parameter unitary group defined by the spectral integral U(t)=∫Reitλ dμ(λ). U(t) = \int_{\mathbb{R}} e^{it\lambda} \, d\mu(\lambda). Then for any t,s∈Rt, s \in \mathbb{R}, the group law holds: U(t+s)=U(t)U(s). U(t+s) = U(t)U(s).

theorem

Strong Continuity of the Unitary Group U(t)=eitTU(t) = e^{itT}

Let HH be a complex Hilbert space and TT a self-adjoint operator with a domain-aware spectral resolution provided by the spectral measure μ\mu on R\mathbb{R}. Let U(t)=eitTU(t) = e^{itT} be the one-parameter unitary group defined by the spectral integral U(t)=∫Reitλ dμ(λ). U(t) = \int_{\mathbb{R}} e^{it\lambda} \, d\mu(\lambda). For any vector x∈Hx \in H, the map t↦U(t)xt \mapsto U(t)x is continuous from R\mathbb{R} to HH.

theorem

Unitary Conjugation of a Spectral Resolution

Let HH and H′H' be complex Hilbert spaces. Suppose TT is a self-adjoint operator on HH with a spectral resolution given by a spectral measure μS\mu_S on R\mathbb{R}. Let u:H→H′u : H \to H' be a unitary operator (a linear isometric equivalence). Then, the unitarily conjugated operator uTu−1u T u^{-1} on H′H' is resolved by the unitarily conjugated spectral measure uμSu−1u \mu_S u^{-1}. This theorem allows for transporting a spectral resolution from one operator to any unitarily equivalent operator.

theorem

Unitary Equivariance of the Domain-Aware Spectral Theorem

Let HH and H′H' be complex Hilbert spaces. Let TT be a partially defined linear operator on HH and μ\mu be a spectral measure on R\mathbb{R}. Suppose that μ\mu provides a domain-aware spectral resolution of TT, meaning that for all xx in the domain of TT and all y∈Hy \in H, the relation ⟨y,Tx⟩=∫Rλ d⟨y,μ(λ)x⟩\langle y, T x \rangle = \int_{\mathbb{R}} \lambda \, d\langle y, \mu(\lambda) x \rangle holds, and the domain of TT is precisely the set of vectors xx for which the second moment is finite: ∫Rλ2 dμx,x<∞\int_{\mathbb{R}} \lambda^2 \, d\mu_{x,x} < \infty. Given a unitary operator u:H→H′u : H \to H', then the unitarily conjugated spectral measure uμu−1u \mu u^{-1} provides a domain-aware spectral resolution for the unitarily conjugated operator uTu−1u T u^{-1}. This implies that the conjugation preserves both the weak spectral integral representation and the characterization of the operator's domain via the square-moment of the spectral measure.