The free Hamiltonian as a self-adjoint operator, and its spectrum #
Category: 1-Mathlib (CSD-free; staged for upstream).
Everything the corpus proves about continuum dynamics is proved with bounded objects: the
unitary group U₀(t) = 𝓕⁻¹ e^{−2π²itξ²} 𝓕 (SchrodingerGroup.lean), the
Fourier multiplier, the kernel. The Hamiltonian itself is unbounded and was never written as an
operator. This module writes it, in the momentum representation, on top of
../InnerProductSpace/MultiplicationOperator.lean:
freeSymbolOp— the free Hamiltonian in the momentum representation: multiplication by the corpus's ownfreeSymbol,m(ξ) = 2π²ξ², onL²(ℝ)with its natural domain, dense (dense_domain_freeSymbolOp);- ★★★
isSelfAdjoint_freeSymbolOp— it is self-adjoint: the first unbounded self-adjoint operator of the continuum in this corpus (the circle's twisted Laplacian,../Fourier/CircleSobolev.lean, is the discrete-spectrum case); - ★★★
spectrum_freeSymbolOp— its spectrum is the nonnegative real axis, both inclusions:⊆because the symbol's values are the nonnegative reals and the resolvent exists off their closure (★range_freeSymbol,isClosed_nonnegAxis), and⊇because every ball around a nonnegative real has preimage of positive measure (it is a nonempty open set) and finite measure (it is bounded,preimage_ball_subset_Icc), so ★★mem_spectrum_freeSymbolOpapplies. The spectrum is purely continuous in the sense that matters here: it is an uncountable set reached by approximate eigenvectors, not by eigenvectors — no eigenvalue is claimed anywhere; coeFn_phaseGroup_freeSymbol— the group and the generator are built from one function: in the momentum representationU₀(t)is multiplication bye^{−itm}for the samem.
Honest scope #
⚠️ The momentum representation. freeSymbolOp is −½Δ conjugated by the Fourier
transform. The position-space operator is its Fourier conjugate, and conjugating a LinearPMap by a
unitary is not built here, so "−½Δ is self-adjoint with spectrum [0, ∞)" is not stated in
position space. The Fourier transform is a unitary of L²(ℝ) (SchrodingerGroup.fourierL2), so the
two operators are unitarily equivalent in the ordinary mathematical sense; that equivalence is what
is missing in Lean (MATHLIB-ABSENT(LinearPMap.conjIsometry)).
⚠️ No H₀ + V. Adding a bounded real potential needs two things this module does not have:
the conjugation above (in momentum space a potential is a convolution, not a multiplier) and the
lemma that a bounded symmetric perturbation of a self-adjoint LinearPMap is self-adjoint on the
same domain. Both are recorded in specs/BACKLOG.md #64 as what the interacting Wigner equation
still waits on.
⚠️ No Stone's theorem, so U₀(t) = e^{−itH₀} is not an identity here: the two objects are
built from the same symbol, which is all coeFn_phaseGroup_freeSymbol says. One dimension.
References: M. Reed, B. Simon, Methods of Modern Mathematical Physics I §VIII.3, II §IX.7;
MultiplicationOperator.lean, SchrodingerGroup.lean, CircleSobolev.lean;
specs/BACKLOG.md #64, #93; specs/future-work.md.
The range of the free symbol #
★ The free symbol's range is [0, ∞).
The nonnegative real axis in ℂ #
The free Hamiltonian in the momentum representation #
The free Hamiltonian in the momentum representation: multiplication by the free symbol
2π²ξ² on L²(ℝ), on its natural domain {f : (1 + ξ²) f ∈ L²}.
Instances For
★★★ The free Hamiltonian is self-adjoint on its natural domain: the first unbounded self-adjoint operator of the continuum in this corpus.
Its spectrum is the nonnegative real axis #
★★ Every nonnegative real is in the spectrum: the preimage of each ball around it is a nonempty open set, hence of positive measure, and bounded, hence of finite measure — so the essential-value criterion applies.
★★★ The spectrum of the free Hamiltonian is the nonnegative real axis [0, ∞): the
continuum spectrum of −½Δ, with no eigenvalues anywhere in it.
The link with the propagator #
The free group, in the momentum representation, is multiplication by e^{−it m} for the same
symbol m whose multiplication operator is the free Hamiltonian: the group and the generator are
built from one function. Turning this into U₀(t) = e^{−itH₀} as operators is Stone's theorem,
which the pin does not have.
The position representation #
#64(i)'s point: with LinearPMap.conjIsometry the momentum-space operator moves across the Fourier
transform, which is a unitary of L²(ℝ) at this pin (MeasureTheory.Lp.fourierTransformₗᵢ). What
comes out is the free Hamiltonian in the position representation — self-adjoint, with the same
spectrum, and defined on the image of the momentum domain.
The free Hamiltonian in the position representation: the momentum-space multiplication operator conjugated by the Fourier transform.
Equations
Instances For
★★ The two representations are unitarily equivalent by construction: applying the position operator to the inverse transform of a momentum-domain state is the inverse transform of the momentum operator's value. This is the statement that makes the name honest.
★★★ The free Hamiltonian is self-adjoint in the position representation too. Self-adjointness is a property of the operator, not of the representation, and this is that sentence as a theorem.
The position-space domain is dense, which it has to be for the operator to be self-adjoint.
★★★ The spectrum is the same nonnegative real axis. The spectrum is representation-
independent, so the continuum spectrum of the free Hamiltonian is [0, ∞) in position space as
well — and the proof is the conjugation lemma, not a second computation.
The Schrödinger operator with a bounded potential #
#64(ii)'s payoff, and the reason (i) and (ii) belong together: the free Hamiltonian lives in the
momentum representation, a potential multiplies in the position representation, so H₀ + V
is only a statement once both are in the same place. (i) moved H₀ across the Fourier transform and
(ii) adds the potential there.
The Schrödinger operator H₀ + V for a bounded real potential, in the position
representation, on the same domain as the free Hamiltonian.
Equations
- SchrodingerGroup.schrodingerOp V hV hb = SchrodingerGroup.freePositionOp.addCLM (MeasureTheory.L2.mulCLM (fun (x : ℝ) => ↑(V x)) hV hb)
Instances For
★★★ The Schrödinger operator is self-adjoint for every bounded real potential, on the free Hamiltonian's own domain. This is what #64's interacting half needed from (i) and (ii) together: the generator of the interacting dynamics exists as a self-adjoint operator, and what is still missing is Stone's theorem to turn it into the unitary group the equation differentiates.
The domain is dense, as it must be.