Differentiation in Fourier coordinates, and the twisted Laplacian on L²(S¹) #
Category: 1-Mathlib (CSD-free; staged for upstream). BACKLOG #93(b)(c), closing #93.
AharonovBohmCircle.lean states the Aharonov–Bohm
eigenvalue equation −(∂ − iΦ)²ψₙ = (n − Φ)²ψₙ pointwise on the Fourier modes, and says honestly
that "spectrum" there means the level set of those modes.
DiagonalOperator.lean supplies the missing operator
theory in general: the diagonal operator of a real weight family on a Hilbert basis is self-adjoint
with spectrum the closure of the weights. This file is the bridge between them, and it needs one
identity that the pin has only in boundary-term form.
Differentiation in Fourier coordinates #
- ★★
fourierCoeffOn_deriv— for aT-periodic function,cₙ(f') = (2πin/T) cₙ(f). The boundary term of Mathlib'sfourierCoeffOn_of_hasDerivAtvanishes by periodicity, and then = 0case is the fundamental theorem of calculus; ★★fourierCoeffOn_deriv_two_piis theT = 2πform, where the factor is exactlyin; fourierCoeffOn_add,fourierCoeffOn_neg— the linearity the pin leaves out;- ★★★
fourierCoeffOn_twistedLap—−(∂ − iΦ)²is multiplication by(n − Φ)²in Fourier coordinates. This is the eigenvalue equation in the form spectral theory needs: not "the modes are eigenvectors" but "the operator is diagonal in the modes".
The operator, its spectrum, and the flux #
twistedOp Φ— the twisted Laplacian onL²(ℝ/2πℤ)asfourierBasis.diagOp (circleEigval Φ). Its domain isH²of the circle: the vectors whose weighted Fourier coefficients are still square-summable, which is whatdiagDomainis;- ★★★
isSelfAdjoint_twistedOpand ★★★spectrum_twistedOp— self-adjoint, with spectrum the closure of the level set, inherited from the generic brick; toL2and ★fourierCoeff_toL2— a continuous periodic function on the line as an element ofL²(ℝ/2πℤ), with the coefficients it came with;- ★★★
twistedOp_apply_of_fourierCoeffand ★★★twistedOp_toL2— the operator acts on everyC²periodic function exactly as−(∂ − iΦ)²does, which is what makes the two theorems above statements about the differential operator and not only about a multiplier; - ★★★
exists_eq_of_spectrum_twistedOp_eq— the flux is determined by the spectrum, modulo a whole quantum and up to sign.AharonovBohmCircle.exists_eq_of_range_circleEigval_eqsaid this of a chosen level set; this says it of the spectrum of a self-adjoint operator onL²(S¹). The extraction isisLeast_norm_image_spectrum: the ground level survives the closure because the bound that produced it is a closed condition.
Honest scope #
⚠️ The identification is on C² periodic functions, not on the whole domain. twistedOp_toL2
covers every twice-differentiable periodic function with continuous second derivative — a dense set
containing all the modes and all trigonometric polynomials — and twistedOp_apply_of_fourierCoeff
covers any pair of L² functions with the right coefficients. What is not proved is that every
element of the domain is twice differentiable in any classical sense: that is a Sobolev embedding
(H²(S¹) ↪ C¹), which the pin does not have and which nothing here needs.
⚠️ No functional calculus and no eigenfunction expansion theorem. The spectrum is computed as a
set; hasSum_fourier_series_L2 is Mathlib's, and is not re-proved or used.
⚠️ The flux quantum here is 1 and the spatial period 2π, as in AharonovBohmCircle.lean; the
lattice ring of AharonovBohmRing.lean is an analogue, not an instance, and nothing here changes
that.
References: the identity cₙ(f') = in·cₙ(f) is the classical one behind the H^k scale — e.g.
Katznelson, An Introduction to Harmonic Analysis I.§4 — and the multiplication-operator model of a
self-adjoint operator with prescribed spectrum is Reed–Simon I, Theorem VIII.4; Y. Aharonov,
D. Bohm, Phys. Rev. 115 (1959) 485; specs/BACKLOG.md #90, #93.
Preliminaries #
The character, read on the line, is continuous.
fourierCoeffOn is additive.
fourierCoeffOn commutes with negation.
The Fourier coefficients of a derivative #
★★ Differentiation multiplies the n-th Fourier coefficient by 2πin/T. The boundary term
of the integration by parts vanishes because the function is periodic; this is the identity the
H^k scale of spaces is built on, and the pin has it only in the boundary-term form
fourierCoeffOn_of_hasDerivAt.
★★ The same at the period 2π, where the factor is exactly in.
The twisted Laplacian in Fourier coordinates #
★★★ The twisted Laplacian is multiplication by (n − Φ)² in Fourier coordinates. This is
the content of the eigenvalue equation of AharonovBohmCircle.lean in the form spectral theory
needs: not "the modes are eigenvectors" but "the operator is diagonal in the modes".
The twisted Laplacian as a self-adjoint operator on L²(ℝ/2πℤ) #
The twisted Laplacian −(∂ − iΦ)² on L²(ℝ/2πℤ): the diagonal operator of the levels
(n − Φ)² in the Fourier basis, on the domain where those weighted coefficients are still
square-summable. That domain is H² of the circle, by definition rather than by construction.
Equations
Instances For
★★★ The twisted Laplacian is self-adjoint on that domain.
★★★ Its spectrum is the closure of the level set — the honest spectrum of a self-adjoint operator, not the set of eigenvalues of a chosen family.
★ Each Fourier mode is an eigenvector with eigenvalue (n − Φ)².
★★★ The operator is the twisted Laplacian. Whenever two L² functions have the Fourier
coefficients that −(∂ − iΦ)² relates — which is what fourierCoeffOn_twistedLap says of a C²
periodic function and its twisted Laplacian — the first is in the domain and the second is its
image. This is the step that makes spectrum_twistedOp a statement about the differential
operator.
From a C² periodic function on the line to the operator #
The derivative of a periodic function is periodic.
A continuous 2π-periodic function on the line, as an element of L²(ℝ/2πℤ).
Equations
- CircleFourier.toL2 f hc hp = (ContinuousMap.toLp 2 AddCircle.haarAddCircle ℂ) { toFun := AddCircle.liftIoc (2 * Real.pi) 0 f, continuous_toFun := ⋯ }
Instances For
★ Its Fourier coefficients are the interval coefficients of the function it came from.
★★★ The operator applied to a C² periodic function is its twisted Laplacian. This closes
the loop: spectrum_twistedOp and isSelfAdjoint_twistedOp are statements about an operator that
acts on every C² periodic function exactly as −(∂ − iΦ)² does.
The flux is determined by the spectrum #
The smallest modulus in the spectrum is the squared distance from the flux to the nearest whole quantum: the ground level survives the closure, because the bound that produced it is closed.
★★★ The flux is determined by the spectrum of the operator, modulo a whole quantum and up to
sign. This is AharonovBohmCircle.exists_eq_of_range_circleEigval_eq upgraded from a statement about
a chosen level set to a statement about the spectrum of a self-adjoint operator on L²(S¹).