Documentation

TauCeti.Analysis.Contour.Argument.CyclePV

The argument principle for a cycle running through the zeros #

The argument principle in the form that tolerates zeros and poles on the contour. For f whose zeros and poles in an open U all lie in a finite S, and a closed piecewise-C¹ immersion γ null-homologous in U and based off S, the Cauchy principal value of the logarithmic integral is

p.v. ∮_γ f'/f = 2πi · Σ_{z ∈ S} n_z(γ) · ord_z f,

with the generalized (non-integer) winding numbers as weights. Where TauCeti.Contour.argumentPrinciple_nullHomologous requires γ to avoid S and produces an ordinary integral, here γ may run through the points of S — a zero on the contour is a pole of f'/f on the contour, so the integral exists only as a principal value, and the point contributes its order weighted by a winding number that need not be an integer. In the standard configuration — a positively oriented curve with a single smooth branch through the point — that weight is 1/2; in general the generalized winding number at such a point is the average of the two ordinary winding numbers on either side of the branch, so any k ± 1/2 occurs.

The proof is the Hungerbühler–Wasem residue theorem in its simple-pole regime (TauCeti.Contour.hungerbuhlerWasem_residueTheorem_of_simple_poles) applied to logDeriv f, whose residue at each point is the order of f there (TauCeti.Contour.residue_logDeriv_eq_meromorphicOrderAt). What that regime asks for is that logDeriv f have at worst a simple pole at each point of S, which is TauCeti.Contour.neg_one_le_meromorphicOrderAt_logDeriv: HW's conditions (A′) and (B) are then automatic, so no regularity hypothesis beyond the immersion survives into the statement. That bound is not an extra assumption on f but a fact about logarithmic derivatives — f'/f has a simple pole at a zero or pole of f whatever the order there — which is why the hypotheses below are those of the classical statement with the avoidance dropped.

Immersion, rather than mere piecewise-C¹ regularity, is what the principal value at an on-curve singularity needs: the curve must leave the point at a definite speed for the excised integrals to converge.

Main results #

Provenance #

Adapted from the AINTLIB LeanModularForms valence-formula development, whose ForMathlib/GeneralizedResidueTheory/Residue.lean and .../Residue/GeneralizedTheoremBase.lean apply the generalized residue theorem to logDeriv f; here stated against Mathlib's meromorphicOrderAt and the raw-γ design of the contour-integration roadmap.

References #

theorem TauCeti.Contour.hasCauchyPV_logDeriv_nullHomologous {f : ℂ → ℂ} {S : Finset ℂ} {U : Set ℂ} {ord : ℂ → ℤ} (hU : IsOpen U) (hoff : ∀ z ∈ U, z ∉ S → AnalyticAt ℂ f z ∧ f z ≠ 0) (hmero : ∀ s ∈ S, s ∈ U → MeromorphicAt f s) (hord : ∀ s ∈ S, s ∈ U → meromorphicOrderAt f s = ↑(ord s)) {γ : ℝ → ℂ} {a b : ℝ} (hγ_imm : IsPwC1ImmersionOn γ a b) (hγU : ∀ t ∈ Set.uIcc a b, γ t ∈ U) (hclosed : γ a = γ b) (hγa : γ a ∉ ↑S) (hnull : IsNullHomologous γ a b U) :
HasCauchyPV γ a b (logDeriv f) (2 * ↑Real.pi * Complex.I * ∑ z ∈ S, windingNumber γ a b z * ↑(ord z))

The argument principle for a cycle running through the zeros. For f analytic and non-vanishing on U off a finite S, meromorphic of order ord at each point of S lying in U, and a closed piecewise-C¹ immersion γ in U, null-homologous there and based off S,

p.v. ∮_γ f'/f = 2πi · Σ_{z ∈ S} n_z(γ) · ord z.

Unlike TauCeti.Contour.argumentPrinciple_nullHomologous, the curve is free to pass through the points of S: where the order is nonzero f'/f then has a pole on the contour, the integral exists only as a principal value, and the generalized winding number supplies the weight, which need not be an integer — 1/2 in the standard configuration of a positively oriented curve with a single smooth branch through the point. Of the curve, only the endpoint γ a is required to avoid S; that restriction is inherited from the Hungerbühler–Wasem theorem, not from the principal value, which excises symmetrically about each singularity.

Nothing is asked of f at a point of S outside U: null-homology forces the winding number there to vanish, so such a term drops out whatever ord says. S may equally list regular non-vanishing points of f, whose order is 0.