The Cayley transform and the spectral theorem for unbounded self-adjoint operators #
The scalar Cayley map C(x) = (x + i)/(x - i) is a bijection of ℝ onto the unit circle
minus the point 1, with inverse D(u) = i (u + 1)/(u - 1). Substituting a self-adjoint
unbounded operator A for x produces a bounded unitary — hence normal — operator U,
to which the spectral theorem for bounded normal operators applies; transporting the
resulting projection-valued measure back along C gives the spectral measure of A.
Main definitions #
cayleyMap,cayleyInv,unitCircleMinusOne— the scalar maps.IsCayleyTransform—Uis the Cayley transform ofA, built from the resolventB = (A - i 1)⁻¹.restrictPVM,mapPVM— restriction and transport of a projection-valued measure.cayleyPVM— the projection-valued measure onℝobtained from that ofU.
Main statements #
unitary_mul_adjoint— a unitary operator is normal.bijOn_cayleyMapand companions — the Cayley map and its inverse.exists_isCayleyTransform— the Cayley transform exists and is unitary,U - 1is injective with rangeDom(A), andA = i (U + 1)(U - 1)⁻¹.mem_spectrum_iff_cayleyMap_mem_spectrum— spectral mapping for the Cayley transform.spectralMeasure_singleton_one_eq_zero—μ^U({1}) = 0.pmapIntegral_cayleyInv_eq—A = ∫ D dμ^U.pmapIntegral_cayleyPVM_id—A = ∫ λ dμ^A.existsUnique_spectralMeasure_unbounded— the spectral theorem for unbounded self-adjoint operators, with concentration on the spectrum.
Unitary operators #
A unitary operator satisfies U* U = U U* = 1; in particular it is normal.
Blueprint reference: lmm:unitary-is-normal.
The scalar Cayley map #
The unit circle in ℂ with the point 1 removed: the image of the real line under the Cayley map.
The spectrum of a unitary operator lies on the unit circle. The Cayley map C sends ℝ
bijectively onto this set (bijOn_cayleyMap), so the spectrum of the Cayley transform of a
self-adjoint operator, with the point 1 removed, corresponds to the spectrum of the operator
(cayleyMap_image_spectrum).
Blueprint: lmm:cayley-map
Instances For
The Cayley map C(x) = (x + i)/(x − i), which sends the real line onto the unit circle with the point 1 removed.
It is the scalar version of the Cayley transform: substituting a self-adjoint operator A for x
gives the unitary operator U = (A + i𝟙)(A − i𝟙)⁻¹. Since |x + i| = |x − i| for real x, C(x) has
modulus 1, and C(x) = 1 is impossible; as x runs from −∞ to +∞, C(x) goes once around the
circle, starting and ending at 1. C is continuous (continuous_cayleyMap) and bijective onto
unitCircleMinusOne (bijOn_cayleyMap), with inverse cayleyInv.
Blueprint: lmm:cayley-map
Instances For
The inverse Cayley map D(u) = i(u + 1)/(u − 1), taken as a real number.
On the unit circle with 1 removed, D is real-valued (cayleyInv_ofReal) and inverts the Cayley
map C (cayleyInv_cayleyMap, cayleyMap_cayleyInv). Integrating D against the spectral
measure of the Cayley transform U recovers the self-adjoint operator A
(pmapIntegral_cayleyInv_eq); this is how the spectral theorem for A is obtained from that of
U.
Implementation note: the value is the real part of i(u + 1)/(u − 1), so that D is a total
function ℂ → ℝ. Off the unit circle minus 1 the value has no meaning (at u = 1 it is 0, by
Lean's convention for division by zero); it is continuous on unitCircleMinusOne
(continuousOn_cayleyInv) and measurable on ℂ (measurable_cayleyInv).
Blueprint: lmm:cayley-map
Instances For
C lands in the unit circle minus 1.
Blueprint reference: lmm:cayley-map.
C is continuous, hence Borel measurable.
Blueprint reference: lmm:cayley-map.
D is continuous on its domain, hence Borel measurable there.
Blueprint reference: lmm:cayley-map.
C ∘ D = id on the unit circle minus 1.
Blueprint reference: lmm:cayley-map.
C is a bijection of ℝ onto the unit circle minus 1.
Blueprint reference: lmm:cayley-map.
The Cayley transform of a self-adjoint operator #
A bounded operator U is the Cayley transform of an operator A, with B the bounded inverse of A − i𝟙: U = (A + i𝟙)(A − i𝟙)⁻¹, that is Uψ = (A + i𝟙)Bψ for every ψ.
The Cayley transform turns an unbounded self-adjoint operator into a bounded unitary one, in the
way the Cayley map turns the real line into the unit circle. For self-adjoint A the inverse B
exists (i is not in the spectrum of A), U is unitary, U − 𝟙 is injective with range Dom(A), and
A can be recovered as i(U + 𝟙)(U − 𝟙)⁻¹ (exists_isCayleyTransform). The spectral theorem for
the bounded normal operator U then yields that of A.
Implementation note: the structure records the inverse B as data alongside U, because the
formula for U applies A to Bψ, which lies in Dom(A). The operator U is determined by A
(IsCayleyTransform.unitary_eq).
Blueprint: thrm:hall-10.28
- isResolvent : IsResolvent A Complex.I B
B is the bounded two-sided inverse of A − i𝟙.
Uψ = (A + i𝟙)Bψ = A(Bψ) + iBψ for every ψ.
Instances For
For a symmetric A and ψ ∈ Dom(A), ‖(A + i 1) ψ‖ = ‖(A - i 1) ψ‖: both squares equal
‖A ψ‖² + ‖ψ‖², the cross terms cancelling by symmetry.
Blueprint reference: thrm:hall-10.28.
Every self-adjoint operator A has a Cayley transform U = (A + i𝟙)(A − i𝟙)⁻¹, and U is unitary.
In detail, with B = (A − i𝟙)⁻¹, which exists because i is not in the spectrum of A:
- U is unitary;
- U − 𝟙 is injective;
- the range of U − 𝟙 is Dom(A), and A(U − 𝟙)χ = i(U + 𝟙)χ for every χ, that is A = i(U + 𝟙)(U − 𝟙)⁻¹ on Dom(A);
- U − 𝟙 = 2iB.
Unitarity rests on ‖(A + i𝟙)ψ‖ = ‖(A − i𝟙)ψ‖ for ψ in Dom(A), which holds because A is
symmetric (norm_add_I_smul_eq_norm_sub_I_smul). Point 3 does not say that 1 is outside the
spectrum of U: (U − 𝟙)⁻¹ is defined only on Dom(A) and is unbounded when A is.
Blueprint: thrm:hall-10.28
Spectral mapping for the Cayley transform: a real λ is in the spectrum of the self-adjoint operator A exactly when C(λ) is in the spectrum of its Cayley transform U.
Together with the reality of the spectrum of A, this gives C(σ(A)) = σ(U) \ {1}
(cayleyMap_image_spectrum), and it is what makes the spectral measure of A, obtained from that
of U, concentrated on σ(A) (spectralMeasure_concentrated).
Implementation note: the statement is about real λ; the spectrum of A is a set of complex
numbers, contained in ℝ by pmapSpectrum_subset_real.
Blueprint: lmm:cayley-spectral-mapping
Consequently C maps σ(A) onto σ(U) \ {1}.
Blueprint reference: lmm:cayley-spectral-mapping.
Transport of projection-valued measures #
The restriction of a projection-valued measure μ on Y to a measurable subset Y₀ whose complement has measure 0: the projection-valued measure on Y₀ sending E to μ(E).
When μ(Y \ Y₀) = 0 no state has any probability of a value outside Y₀, so nothing is lost by regarding μ as a measure on Y₀; in particular μ(Y₀) = 𝟙.
Implementation note: Y₀ is a subtype, with the subspace σ-algebra; a set E of Y₀ goes to μ of
its image in Y. The construction of cayleyPVM does not use this definition.
Blueprint: lmm:borel-bijection-transports-pvm (Part 1)
Equations
- One or more equations did not get rendered due to their size.
Instances For
The transport of a projection-valued measure μ on Y along a measurable bijection T : Y → Z with measurable inverse: the projection-valued measure on Z sending F to μ(T⁻¹(F)).
This relabels the possible values of an observable by T without changing the projections. The
measure it assigns to a vector ψ is the pushforward along T of the one μ assigns
(assoc_mapPVM).
Implementation note: measurability of the inverse ensures that T⁻¹(F) is non-measurable
whenever F is, so that the transported measure is 0 on non-measurable sets, as
ProjectionValuedMeasure requires. pushforwardPVM is the variant for a map that is not
bijective. The construction of cayleyPVM does not use this definition.
Blueprint: lmm:borel-bijection-transports-pvm (Part 2)
Equations
Instances For
The scalar measures of the transported projection-valued measure are the pushforwards.
Blueprint reference: lmm:borel-bijection-transports-pvm (Part 2).
The spectral measure of an unbounded self-adjoint operator #
μ^U assigns no mass to {1}, so D — undefined there — is defined almost everywhere
for every μ^U_ψ.
Blueprint reference: lmm:cayley-omits-one.
D is Borel measurable on all of ℂ.
A self-adjoint operator has no proper self-adjoint extension.
The Cayley transform U is determined by A.
For unitary U with μ^U({1}) = 0 and ψ = (U - 1) χ: ψ ∈ W_D and
(∫ D dμ^U) ψ = i (U + 1) χ, since D(u) (u - 1) = i (u + 1) off u = 1.
A is recovered from U by integrating D against μ^U, with equality of domains.
Blueprint reference: prpstn:hall-10.29.
The projection-valued measure on ℝ obtained from a projection-valued measure μ(U) on the spectrum of a unitary U through the inverse Cayley map: a Borel set E goes to μ(U)(C(E)).
When U is the Cayley transform of a self-adjoint A and μ(U) its spectral measure, this is the
spectral measure of A: ∫ λ dμ(λ) = A (pmapIntegral_cayleyPVM_id). The point 1 of the circle,
which corresponds to no real number, carries no mass (spectralMeasure_singleton_one_eq_zero),
so nothing of μ(U) is lost.
Implementation note: C(E) is written as the set of u in σ(U) with D(u) in E, using
cayleyInv; the two agree on the unit circle minus 1. Non-measurable E are sent to 0, as a
projection-valued measure requires; the preimage of a non-Borel set under D can still carry
mass. The blueprint builds the measure by restriction and transport (restrictPVM,
mapPVM); this definition composes with D directly.
Blueprint: thrm:hall-10.30
Equations
- One or more equations did not get rendered due to their size.
Instances For
The scalar measures of μ^A are the pushforwards of those of μ^U along D.
∫ λ dμ^A = ∫ D dμ^U, with equality of domains, by change of variables.
Transporting the spectral measure of U through the Cayley map gives a projection-valued
measure on ℝ integrating the identity to A.
Blueprint reference: thrm:hall-10.30.
The pushforward of a projection-valued measure μ on Y along a measurable map T : Y → Z: the projection-valued measure on Z sending a measurable F to μ(T⁻¹(F)).
For a bounded measurable g on Z, ∫ g dμ' = ∫ (g ∘ T) dμ, where μ' is the pushforward
(integral_pushforwardPVM). The uniqueness half of the unbounded spectral theorem uses it to
move a second spectral measure ν of A, on ℝ, to the circle along the Cayley map, where it is
compared with the spectral measure of the Cayley transform U.
Implementation note: T need not be bijective, unlike mapPVM; in exchange non-measurable F are
sent to 0 by an explicit case split.
Blueprint: thrm:hall-10.4 (formalization note after its proof)
Equations
- One or more equations did not get rendered due to their size.
Instances For
Change of variables for the bounded integral: ∫ g d(T_* μ) = ∫ g ∘ T dμ.
If ∫ λ dμ = A and B = (A - i 1)⁻¹, then ∫ (λ - i)⁻¹ dμ = B.
Blueprint reference: thrm:hall-10.4 (Step 1 of the uniqueness proof).
If ∫ λ dμ = A, then ∫ C dμ = U for the Cayley map C and the Cayley transform U
of A, since C = 1 + 2i (λ - i)⁻¹ and U = 1 + 2i (A - i 1)⁻¹.
Blueprint reference: thrm:hall-10.4 (Step 2 of the uniqueness proof).
Uniqueness: any projection-valued measure on ℝ integrating the identity to A is the
Cayley transport μ^A of the spectral measure μ^U of the Cayley transform U.
Blueprint reference: thrm:hall-10.4 (Steps 3 and 4 of the uniqueness proof).
The spectral theorem for unbounded self-adjoint operators: for every self-adjoint A exactly one projection-valued measure μ on ℝ satisfies ∫ λ dμ(λ) = A.
This is the mathematical content of the statement that an observable, such as position,
momentum or a Hamiltonian, has a spectral decomposition: μ(E) is the projection onto the
states where the observable takes a value in E with certainty, and ⟨ψ, μ(E)ψ⟩ is the
probability of a value in E for the unit vector ψ. The integral is the unbounded operator of
pmapIntegral, domain included, so the equation also says that Dom(A) consists of the ψ with
∫ λ² dμ(ψ)(λ) finite, where μ(ψ) is the measure E ↦ ⟨ψ, μ(E)ψ⟩. The measure is concentrated on the
spectrum of A (spectralMeasure_concentrated).
Implementation note: existence is cayleyPVM of the spectral measure of the Cayley transform
of A (exists_isCayleyTransform, existsUnique_spectralMeasure_normal); uniqueness passes
through eq_cayleyPVM_of_pmapIntegral_eq. The case H = {0} is treated separately. The measure
itself is spectralMeasure.
Blueprint: thrm:hall-10.4
The spectral measure is concentrated on the spectrum.
Blueprint reference: thrm:hall-10.4.