Physlib

PhyslibAlpha.ProbabilisticTheory.HilbertSpace.Unbounded.UnitaryInfra.SpectralMeasure

The spectral measure of a bounded normal operator

i. Overview

The sesquilinear form of the indicator of a Borel set `S` is represented by an operator `E(S)`. These operators are projections, countably additive in the weak operator topology, and commute with every operator commuting with `U` and `U⋆`. They form the spectral measure of `U`, whose integral of the identity is `U`.

ii. Key results

  • `cfcSpectralOperator` : the operator `E(S)`.
  • `cfcSpectralMeasure` : the spectral measure of a bounded normal operator.

iii. Table of contents

  • A. Order properties of the spectral operators
  • B. Weak countable additivity
  • C. The spectral operators are projections
  • D. The spectral measure
  • E. Commutation with the commutant

A. Order properties of the spectral operators

B. Weak countable additivity

C. The spectral operators are projections

For a compact set `K` and an open `V ⊇ K`, a Urysohn function `f` gives `E(K) ≤ f(U) ≤ E(V)`. With the estimate `‖A x‖² ≤ re ⟪x, A x⟫` for `0 ≤ A ≤ 1` and outer regularity this shows `E(K)² = E(K)`. Inner regularity extends it to measurable sets, and multiplicativity on intersections follows from additivity.

The idempotency proof above is the reusable order/regularity construction. It first proves the compact case with an Urysohn cutoff and outer regularity, then approximates an arbitrary measurable set from inside by compact sets. The only analytic estimate is the positive contraction inequality `0 ≤ A ≤ 1 ⟹ ‖A x‖² ≤ re ⟪x, A x⟫`; no weak/strong operator topology argument or multiplication of operator limits is involved.

D. The spectral measure

E. Commutation with the commutant

An operator commuting with `U` and `U⋆` commutes with `f(U)` for continuous `f`, and by uniqueness in the Riesz–Markov–Kakutani theorem with every spectral projection of `U`.

48 declarations

definition

Spectral operator E(S)E(S) of a bounded normal operator UU

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For a measurable set S⊆σ(U)S \subseteq \sigma(U) in the spectrum of UU, `cfcSpectralOperatorAux` is the continuous linear operator E(S):H→HE(S): H \to H obtained via the Riesz representation theorem from the sesquilinear form BS(x,y)=μx,y(S)‾B_S(x, y) = \overline{\mu_{x, y}(S)}. It is the unique operator such that for any x,y∈Hx, y \in H, the inner product ⟨y,E(S)x⟩\langle y, E(S) x \rangle is equal to the value of the polarized complex scalar measure μx,y\mu_{x, y} evaluated at SS.

theorem

The Inner Product ⟨y,E(S)x⟩\langle y, E(S) x \rangle Equals the Polarized Measure μx,y(S)\mu_{x,y}(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any Borel measurable set S⊆σ(U)S \subseteq \sigma(U) and any vectors x,y∈Hx, y \in H, the inner product of the spectral operator E(S)E(S) satisfies ⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S) where μx,y\mu_{x, y} is the polarized complex scalar measure associated with the operator UU and vectors x,yx, y.

definition

Spectral operator E(S)E(S) of a bounded normal operator UU

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S):H→HE(S): H \to H is defined as follows: if SS is a measurable set, E(S)E(S) is the unique continuous linear operator such that for all x,y∈Hx, y \in H, the inner product ⟨y,E(S)x⟩\langle y, E(S) x \rangle equals the value of the polarized complex scalar measure μx,y\mu_{x, y} evaluated at SS. If SS is not measurable, E(S)E(S) is defined to be the zero operator 00.

theorem

⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any subset S⊆σ(U)S \subseteq \sigma(U) and any vectors x,y∈Hx, y \in H, the inner product of yy with the spectral operator E(S)E(S) applied to xx is equal to the polarized complex scalar measure μx,y\mu_{x, y} evaluated at SS: ⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S)

theorem

The spectral operator E(S)E(S) is zero for non-measurable sets SS

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let σ(U)\sigma(U) denote the spectrum of UU. For any subset S⊆σ(U)S \subseteq \sigma(U), if SS is not a measurable set, then the spectral operator E(S)E(S) associated with UU is the zero operator 00.

theorem

The scalar spectral measure μv\mu_v equals μ−iv\mu_{-iv}

Let UU be a bounded normal operator on a complex Hilbert space HH and σ(U)\sigma(U) be its spectrum. For any vector v∈Hv \in H, let μv\mu_v be the scalar spectral measure on σ(U)\sigma(U) associated with the vector vv (the regular Borel measure representing the functional f↦Re⟨v,f(U)v⟩f \mapsto \text{Re} \langle v, f(U) v \rangle for f∈C(σ(U),R)f \in C(\sigma(U), \mathbb{R})). Then for any Borel set S⊆σ(U)S \subseteq \sigma(U), the measure of SS with respect to vv is equal to the measure of SS with respect to the vector rotated by the negative imaginary unit, −iv-iv: μv(S)=μ−iv(S)\mu_v(S) = \mu_{-iv}(S)

theorem

Conjugate Symmetry of the Polarized Spectral Measure: μy,x(S)=μx,y(S)‾\mu_{y,x}(S) = \overline{\mu_{x,y}(S)}

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any two vectors x,y∈Hx, y \in H and any Borel set S⊆σ(U)S \subseteq \sigma(U), let μx,y\mu_{x, y} be the polarized complex spectral measure associated with UU. Then the measure satisfies the conjugate symmetry property: μy,x(S)=μx,y(S)‾ \mu_{y, x}(S) = \overline{\mu_{x, y}(S)} where z‾\overline{z} denotes the complex conjugate of z∈Cz \in \mathbb{C}. This identity reflects the Hermitian symmetry of the underlying sesquilinear form, matching the convention that the spectral measure represents the inner product ⟨y,E(S)x⟩\langle y, E(S) x \rangle.

theorem

The spectral operator E(S)E(S) is self-adjoint

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any subset S⊆σ(U)S \subseteq \sigma(U), the spectral operator E(S)E(S) (formally `cfcSpectralOperator U hU S`) is self-adjoint. If SS is measurable, E(S)E(S) is the unique operator such that ⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S) for all x,y∈Hx, y \in H, where μx,y\mu_{x, y} is the polarized complex spectral measure; if SS is not measurable, E(S)=0E(S) = 0. In both cases, E(S)E(S) satisfies the property ⟨y,E(S)x⟩=⟨E(S)y,x⟩\langle y, E(S) x \rangle = \langle E(S) y, x \rangle.

theorem

μx,y(σ(U))=⟨y,x⟩\mu_{x, y}(\sigma(U)) = \langle y, x \rangle

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let μx,y\mu_{x, y} denote the polarized complex spectral measure associated with vectors x,y∈Hx, y \in H, which is a measure defined on the spectrum σ(U)\sigma(U). Then the measure of the whole spectrum is equal to the inner product of the vectors: μx,y(σ(U))=⟨y,x⟩ \mu_{x, y}(\sigma(U)) = \langle y, x \rangle

theorem

E(σ(U))x=xE(\sigma(U)) x = x

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let E(S)E(S) denote the spectral operator associated with a subset SS of the spectrum σ(U)\sigma(U). Then for any vector x∈Hx \in H, the spectral operator associated with the entire spectrum applied to xx is equal to xx, i.e., E(σ(U))x=xE(\sigma(U))x = x.

theorem

E(σ(U))=IE(\sigma(U)) = I

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let E(S)E(S) denote the spectral operator (as defined by the continuous functional calculus) for a subset SS of the spectrum σ(U)\sigma(U). Then the spectral operator evaluated on the entire spectrum is the identity operator II on HH: E(σ(U))=I E(\sigma(U)) = I

theorem

⟨x,E(S)x⟩=μx(S)\langle x, E(S) x \rangle = \mu_x(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let σ(U)\sigma(U) denote the spectrum of UU. For any measurable set S⊆σ(U)S \subseteq \sigma(U) and any vector x∈Hx \in H, the inner product of xx and the spectral operator E(S)E(S) applied to xx is equal to the value of the scalar spectral measure μx\mu_x evaluated at SS: ⟨x,E(S)x⟩=μx(S)\langle x, E(S) x \rangle = \mu_x(S) Here, E(S)E(S) (represented by `cfcSpectralOperator`) is the spectral projection associated with the set SS, and μx\mu_x (represented by `cfcScalarMeasure`) is the regular Borel measure on the spectrum associated with the vector xx via the Riesz–Markov–Kakutani representation theorem.

theorem

Re⟨x,E(S)x⟩=μx(S)\text{Re} \langle x, E(S)x \rangle = \mu_x(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU and any vector x∈Hx \in H, the real part of the quadratic form of the spectral operator E(S)E(S) at xx is equal to the value of the real-valued scalar spectral measure μx\mu_x evaluated at SS. That is, Re⟨x,E(S)x⟩=μx(S)\text{Re} \langle x, E(S) x \rangle = \mu_x(S) where E(S)E(S) is the spectral operator associated with SS and μx\mu_x is the regular Borel measure on σ(U)\sigma(U) associated with xx.

theorem

The spectral operator E(S)E(S) is positive for measurable SS

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) is a positive operator, meaning that ⟨x,E(S)x⟩≥0\langle x, E(S) x \rangle \ge 0 for all x∈Hx \in H.

theorem

The spectral operator E(S)E(S) is non-negative (0≤E(S)0 \le E(S))

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) is non-negative, i.e., 0≤E(S)0 \le E(S).

theorem

Monotonicity of the spectral operator E(S)E(S) with respect to set inclusion

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let σ(U)\sigma(U) denote the spectrum of UU. For any measurable subsets S,T⊆σ(U)S, T \subseteq \sigma(U), if S⊆TS \subseteq T, then the corresponding spectral operators satisfy E(S)≤E(T)E(S) \le E(T) in the Loewner order (i.e., E(T)−E(S)E(T) - E(S) is a positive operator).

theorem

Spectral operator E(S)≤IE(S) \le I for measurable SS

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) satisfies E(S)≤IE(S) \le I, where II denotes the identity operator on HH.

theorem

The spectral operator of the empty set is zero (E(∅)=0E(\emptyset) = 0)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let E(S)E(S) denote the spectral operator (provided by the continuous functional calculus) associated with a subset SS of the spectrum σ(U)\sigma(U). Then, the spectral operator of the empty set is the zero operator, i.e., E(∅)=0.E(\emptyset) = 0.

theorem

E(S∪T)=E(S)+E(T)E(S \cup T) = E(S) + E(T) for disjoint measurable sets SS and TT

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let E(S)E(S) denote the spectral operator associated with UU for a subset S⊆σ(U)S \subseteq \sigma(U). For any two disjoint measurable subsets SS and TT of the spectrum σ(U)\sigma(U), the spectral operator of their union is the sum of their individual spectral operators: E(S∪T)=E(S)+E(T)E(S \cup T) = E(S) + E(T)

definition

Spectral operator E(S)E(S) as a WOT-valued function

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, this function returns the spectral operator E(S)E(S) (see `cfcSpectralOperator`) as an element of the space of bounded linear operators on HH equipped with the weak operator topology (WOT).

theorem

⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any subset S⊆σ(U)S \subseteq \sigma(U) and any vectors x,y∈Hx, y \in H, the inner product of yy with the spectral operator E(S)E(S) acting on xx is equal to the value of the polarized complex spectral measure μx,y\mu_{x, y} evaluated on the set SS: ⟨y,E(S)x⟩=μx,y(S)\langle y, E(S) x \rangle = \mu_{x, y}(S) where E(S)E(S) (represented by `cfcSpectralMeasureFun`) is the projection-valued measure associated with UU, and μx,y\mu_{x, y} is the polarized complex spectral measure corresponding to the vectors xx and yy.

definition

Spectral vector measure of a bounded normal operator UU

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. The spectral vector measure is a vector measure EE that assigns to each Borel subset SS of the spectrum σ(U)\sigma(U) a bounded linear operator E(S)E(S) on HH. This measure takes values in the space of bounded linear operators equipped with the weak operator topology (WOT). Specifically, for any vectors x,y∈Hx, y \in H, the set function S↦⟨y,E(S)x⟩S \mapsto \langle y, E(S) x \rangle is a complex-valued Borel measure on σ(U)\sigma(U), and the operator-valued measure is countably additive in the weak sense.

theorem

The spectral vector measure evaluated at SS is equal to E(S)E(S)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any subset S⊆σ(U)S \subseteq \sigma(U), the spectral vector measure evaluated at SS is equal to the spectral operator E(S)E(S) defined by the spectral measure function.

theorem

μ(K)≤∫f dμ\mu(K) \le \int f \, d\mu for non-negative ff equal to 11 on compact KK

Let μ\mu be a finite measure on a space XX. For any compact set K⊆XK \subseteq X and any non-negative continuous function f:X→Rf: X \to \mathbb{R} with compact support such that f(x)=1f(x) = 1 for all x∈Kx \in K, the measure of KK is bounded by the integral of ff: μ(K)≤∫Xf dμ\mu(K) \le \int_X f \, d\mu

theorem

∫f dμ≤μ(V)\int f \, d\mu \le \mu(V) for f∈[0,1]f \in [0, 1] with supp(f)⊆V\text{supp}(f) \subseteq V

Let μ\mu be a finite measure on a space XX. Let V⊆XV \subseteq X be an open set and f:X→Rf: X \to \mathbb{R} be a continuous function with compact support. If the support of ff is contained in VV (i.e., supp(f)⊆V\text{supp}(f) \subseteq V) and 0≤f(x)≤10 \le f(x) \le 1 for all x∈Xx \in X, then the integral of ff with respect to μ\mu satisfies ∫Xf dμ≤μ(V),\int_X f \, d\mu \leq \mu(V), where μ(V)\mu(V) denotes the measure of the set VV.

theorem

∫f dμ≤μ(D)\int f \, d\mu \leq \mu(D) for functions ff bounded by the indicator 1D\mathbf{1}_D

Let XX be a space with a finite measure μ\mu. Let D⊆XD \subseteq X be a measurable set and f:X→Rf: X \to \mathbb{R} be a non-negative continuous function with compact support. If f(x)≤1D(x)f(x) \leq \mathbf{1}_D(x) for all x∈Xx \in X (where 1D\mathbf{1}_D is the indicator function of DD), then the integral of ff with respect to μ\mu is less than or equal to the measure of DD: ∫Xf dμ≤μ(D)\int_X f \, d\mu \leq \mu(D)

theorem

μ(S)<ϵ  ⟹  μR(S)<ϵ\mu(S) < \epsilon \implies \mu_{\mathbb{R}}(S) < \epsilon for finite measures

Let μ\mu be a finite measure on a space XX. For any subset S⊆XS \subseteq X and any real number ϵ>0\epsilon > 0, if the measure μ(S)\mu(S) is strictly less than ϵ\epsilon (expressed as an extended non-negative real number), then the real-valued measure of SS, denoted μR(S)\mu_{\mathbb{R}}(S), is also strictly less than ϵ\epsilon.

theorem

The spectral operator E(S)E(S) is non-negative

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) is non-negative, meaning that 0≤E(S)0 \le E(S).

theorem

The spectral operator E(S)E(S) is less than or equal to the identity operator II

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable subset SS of the spectrum σ(U)\sigma(U), the spectral operator E(S)E(S) satisfies E(S)≤IE(S) \le I, where II is the identity operator on HH.

theorem

S⊆T  ⟹  E(S)≤E(T)S \subseteq T \implies E(S) \le E(T)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any two measurable sets S,T⊆σ(U)S, T \subseteq \sigma(U), if S⊆TS \subseteq T, then the corresponding spectral operators E(S)E(S) and E(T)E(T) (defined via the continuous functional calculus) satisfy the operator inequality E(S)≤E(T)E(S) \le E(T). This means that for all x∈Hx \in H, the inner products satisfy ⟨x,E(S)x⟩≤⟨x,E(T)x⟩\langle x, E(S) x \rangle \le \langle x, E(T) x \rangle.

theorem

Re⟨x,f(U)x⟩=∫f dμx\text{Re} \langle x, f(U) x \rangle = \int f \, d\mu_x

Let UU be a bounded normal operator on a complex Hilbert space HH, and let σ(U)\sigma(U) denote its spectrum. For any real-valued continuous function ff with compact support on σ(U)\sigma(U) and any vector x∈Hx \in H, the real part of the inner product ⟨x,f(U)x⟩\langle x, f(U) x \rangle is equal to the integral of ff with respect to the scalar spectral measure μx\mu_x associated with xx: Re⟨x,f(U)x⟩=∫σ(U)f(λ) dμx(λ)\text{Re} \langle x, f(U) x \rangle = \int_{\sigma(U)} f(\lambda) \, d\mu_x(\lambda)

theorem

E(K)≤f(U)E(K) \le f(U) for a continuous function ff that is 11 on the compact set KK

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). Let K⊆σ(U)K \subseteq \sigma(U) be a compact set and V⊆σ(U)V \subseteq \sigma(U) be an open set such that K⊆VK \subseteq V. Let f:σ(U)→Rf: \sigma(U) \to \mathbb{R} be a continuous function with compact support such that 0≤f(z)≤10 \le f(z) \le 1 for all z∈σ(U)z \in \sigma(U). If f(z)=1f(z) = 1 for all z∈Kz \in K and the support of ff is contained in VV, then the spectral operator E(K)E(K) and the operator f(U)f(U) (obtained via the continuous functional calculus) satisfy the operator inequality: E(K)≤f(U)E(K) \le f(U) This inequality implies that for all x∈Hx \in H, the inner products satisfy ⟨x,E(K)x⟩≤⟨x,f(U)x⟩\langle x, E(K) x \rangle \le \langle x, f(U) x \rangle.

theorem

f(U)≤E(V)f(U) \le E(V) when 0≤f≤10 \le f \le 1 and supp(f)⊆V\text{supp}(f) \subseteq V

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). Let V⊆σ(U)V \subseteq \sigma(U) be an open set. For any real-valued, compactly supported continuous function f:σ(U)→Rf: \sigma(U) \to \mathbb{R} such that 0≤f(λ)≤10 \le f(\lambda) \le 1 for all λ∈σ(U)\lambda \in \sigma(U) and whose topological support supp(f)\text{supp}(f) is contained in VV, the operator f(U)f(U) defined by the continuous functional calculus and the spectral projection operator E(V)E(V) satisfy the operator inequality: f(U)≤E(V)f(U) \le E(V) In this context, A≤BA \le B denotes the standard Loewner order on self-adjoint operators, meaning ⟨x,Ax⟩≤⟨x,Bx⟩\langle x, Ax \rangle \le \langle x, Bx \rangle for all x∈Hx \in H.

theorem

Multiplicativity of the functional calculus for f2f^2

Let UU be a bounded normal operator on a complex Hilbert space HH, and let σ(U)\sigma(U) denote its spectrum. For any real-valued, compactly supported continuous function f:σ(U)→Rf: \sigma(U) \to \mathbb{R}, let f(U)f(U) denote the operator obtained via the continuous functional calculus. Then the operator corresponding to the pointwise product f2f^2 satisfies (f2)(U)=f(U)2(f^2)(U) = f(U)^2.

theorem

(f−f2)(U)=f(U)−f(U)2(f - f^2)(U) = f(U) - f(U)^2 for the continuous functional calculus

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). For any continuous real-valued function with compact support f∈Cc(σ(U),R)f \in C_c(\sigma(U), \mathbb{R}), the operator (f−f2)(U)(f - f^2)(U) obtained via the continuous functional calculus satisfies the identity: (f−f2)(U)=f(U)−f(U)2(f - f^2)(U) = f(U) - f(U)^2 where f(U)f(U) denotes the continuous functional calculus of UU associated with the function ff.

theorem

The Spectral Operator E(K)E(K) is Idempotent for any Compact Set KK

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let σ(U)\sigma(U) denote the spectrum of UU, and for any subset K⊆σ(U)K \subseteq \sigma(U), let E(K)E(K) denote the spectral operator associated with KK. If KK is a compact set, then E(K)E(K) is idempotent, satisfying the identity E(K)2=E(K)E(K)^2 = E(K) where the multiplication denotes the composition of operators.

theorem

The Spectral Operator E(S)E(S) is Idempotent for any Measurable Set SS

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any measurable set S⊆σ(U)S \subseteq \sigma(U) in the spectrum of UU, let E(S)E(S) denote the spectral operator (formally `cfcSpectralOperatorAux`). Then E(S)E(S) is idempotent, satisfying the identity E(S)2=E(S)E(S)^2 = E(S) where the multiplication denotes the composition of operators.

theorem

The spectral operator E(S)E(S) is idempotent for any set S⊆σ(U)S \subseteq \sigma(U)

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, let E(S)E(S) denote the spectral operator (formally `cfcSpectralOperator`) associated with SS. Then E(S)E(S) is an idempotent operator, satisfying the identity E(S)2=E(S)E(S)^2 = E(S) where the multiplication denotes the composition of operators. If SS is not a measurable set, E(S)E(S) is defined to be the zero operator, which is trivially idempotent.

theorem

The spectral operator E(S)E(S) is a star-projection

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. For any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) (formally `cfcSpectralOperator U hU S`) is a star-projection. This means that E(S)E(S) is both self-adjoint, satisfying E(S)∗=E(S)E(S)^* = E(S), and idempotent, satisfying E(S)2=E(S)E(S)^2 = E(S).

definition

Spectral measure of a bounded normal operator UU

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. The spectral measure of UU is a mapping EE from the Borel subsets of the spectrum σ(U)\sigma(U) to the space of bounded linear operators on HH equipped with the weak operator topology. For any Borel set S⊆σ(U)S \subseteq \sigma(U), the operator E(S)E(S) is a star-projection, meaning it is both idempotent (E(S)2=E(S)E(S)^2 = E(S)) and self-adjoint (E(S)∗=E(S)E(S)^* = E(S)). This measure is countably additive in the weak operator topology and satisfies E(σ(U))=IE(\sigma(U)) = I, where II is the identity operator on HH.

theorem

Evaluation of the Spectral Measure E(S)E(S) at a Set SS

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let EE denote the spectral measure of UU defined on its spectrum σ(U)\sigma(U). For any subset S⊆σ(U)S \subseteq \sigma(U), the value of the spectral measure EE evaluated at SS is equal to the spectral operator E(S)E(S) (represented by `cfcSpectralMeasureFun`).

theorem

The scalar measure ⟨y,E(⋅)x⟩\langle y, E(\cdot) x \rangle equals the polarized spectral measure μx,y\mu_{x, y}

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). Let EE be the spectral measure of UU (denoted by `cfcSpectralMeasure`). For any two vectors x,y∈Hx, y \in H, the scalar complex measure defined by S↦⟨y,E(S)x⟩S \mapsto \langle y, E(S)x \rangle for Borel sets S⊆σ(U)S \subseteq \sigma(U) is equal to the polarized complex spectral measure μx,y\mu_{x, y} associated with xx and yy. (This measure μx,y\mu_{x, y} is the unique complex measure such that ∫σ(U)f dμx,y=⟨y,f(U)x⟩\int_{\sigma(U)} f \, d\mu_{x, y} = \langle y, f(U)x \rangle for every continuous function ff on the spectrum).

theorem

Spectral measure reconstruction of the continuous functional calculus: ⟨y,f(U)x⟩=∫f d⟨y,Ex⟩\langle y, f(U) x \rangle = \int f \, d\langle y, E x \rangle

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH with spectrum σ(U)\sigma(U). Let EE be the spectral measure associated with UU. For any continuous real-valued function f:σ(U)→Rf: \sigma(U) \to \mathbb{R} and any vectors x,y∈Hx, y \in H, the integral of ff with respect to the complex-valued scalar measure μx,y(S)=⟨y,E(S)x⟩\mu_{x, y}(S) = \langle y, E(S) x \rangle is equal to the inner product of yy with the operator f(U)f(U) acting on xx. That is, ∫σ(U)f(λ) d⟨y,E(λ)x⟩=⟨y,f(U)x⟩,\int_{\sigma(U)} f(\lambda) \, d\langle y, E(\lambda) x \rangle = \langle y, f(U) x \rangle, where f(U)f(U) is the operator defined by the continuous functional calculus for UU.

theorem

TT commutes with f(U)f(U) if TT commutes with UU and U∗U^*

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let TT be a bounded linear operator on HH. If TT commutes with UU (i.e., UT=TUUT = TU) and TT commutes with the adjoint operator U∗U^* (i.e., U∗T=TU∗U^*T = TU^*), then for any continuous real-valued function ff defined on the spectrum σ(U)\sigma(U), TT commutes with the operator f(U)f(U) obtained via the continuous functional calculus.

theorem

μTv=μv\mu_{Tv} = \mu_v for an isometric intertwiner TT of {U,U∗}\{U, U^*\}

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let TT be a bounded linear operator on HH that is an isometry (i.e., T∗T=IT^* T = I) and commutes with both UU and its adjoint U∗U^*. Then for any vector v∈Hv \in H, the scalar spectral measure μTv\mu_{Tv} of UU associated with the vector TvTv is equal to the scalar spectral measure μv\mu_v associated with vv.

theorem

μTx,y=μx,T∗y\mu_{Tx, y} = \mu_{x, T^* y} for unitary operators commuting with UU and U∗U^*

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let TT be a unitary operator on HH that commutes with both UU and its adjoint U∗U^* (i.e., UT=TUUT = TU and U∗T=TU∗U^*T = TU^*). For any two vectors x,y∈Hx, y \in H, the polarized complex spectral measure μx,y\mu_{x, y} associated with UU satisfies the identity: μTx,y=μx,T∗y \mu_{Tx, y} = \mu_{x, T^* y} where T∗T^* denotes the adjoint of TT. This indicates that the sesquilinear form of the spectral measure intertwines with the unitary operator TT such that testing at (Tx,y)(Tx, y) is equivalent to testing at (x,T∗y)(x, T^* y).

theorem

The spectral operator E(S)E(S) commutes with unitary operators commuting with UU and U∗U^*

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let TT be a unitary operator on HH that commutes with both UU and its adjoint U∗U^*, such that UT=TUUT = TU and U∗T=TU∗U^*T = TU^*. Then, for any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral operator E(S)E(S) commutes with TT: E(S)T=TE(S) E(S)T = TE(S) This property holds for any set SS, regardless of whether it is measurable.

theorem

The Spectral Measure E(S)E(S) Commutes with Unitary Intertwiners of UU

Let HH be a complex Hilbert space and UU be a bounded normal operator on HH. Let TT be a unitary operator on HH that commutes with both UU and its adjoint U∗U^* (i.e., UT=TUUT = TU and U∗T=TU∗U^*T = TU^*). Then, for any subset S⊆σ(U)S \subseteq \sigma(U) of the spectrum of UU, the spectral measure E(S)E(S) (viewed in the weak operator topology) commutes with TT: E(S)T=TE(S) E(S)T = TE(S) This property holds for any set SS, regardless of whether it is measurable.