Documentation

TauCeti.Analysis.Sobolev.Circle.Neck

The neck transfer of a plumbing node on the circle Sobolev spaces #

The neck of the plumbing xy = a, 0 < |a| < 1, is the annulus |a| ≤ |x| ≤ 1 in the first branch coordinate, the second branch coordinate being y = a / x. Its two outer boundary circles are |x| = 1 and |y| = 1. A function G(y) = ∑ₙ Gₙ yⁿ of the second coordinate, read on the circle |x| = 1, becomes x ↦ G(a / x), whose (-n)-th Fourier coefficient is aⁿ Gₙ.

This file studies the resulting operator on the circle Sobolev spaces H^s. The neck transfer CircleSobolev.neckTransfer a sends u to ∑_{n > 0} aⁿ uₙ e^{-inθ}: it carries the positive frequencies on one boundary circle to the negative frequencies they produce on the other. For ‖a‖ ≤ 1 it has norm at most ‖a‖, it depends holomorphically on a in the open unit disc, and its derivative at a = 0 is the operator u ↦ u₁ e^{-iθ}, which is nonzero for E ≠ 0.

The linearized node equation F(x) + G(a / x) = const on the neck reads, in the Fourier coefficients on the two outer circles, F₋ₙ = -aⁿ Gₙ and G₋ₙ = -aⁿ Fₙ for every n > 0. The pairs (F, G) satisfying it form CircleSobolev.neckGraph s E a. For ‖a‖ ≤ 1 their negative frequency parts are -neckTransfer a G and -neckTransfer a F, so the neck graph is the graph, over the pair of nonnegative Laurent subspaces, of a bounded operator that vanishes at a = 0. It is a closed complement of the pair of negative Laurent subspaces, and at a = 0 it is the pair H^s_{≥0} × H^s_{≥0} of boundary values of holomorphic functions on the two branch discs of the node xy = 0. This is the linear model of the Hardy decomposition at a node.

Main definitions #

Main statements #

References #

The neck transfer #

The neck transfer of the plumbing xy = a: the operator on H^s sending u to ∑_{n > 0} aⁿ uₙ e^{-inθ}. For s > 1/2 and complete E, on a nonnegative-frequency u, the boundary value of a holomorphic function U on the unit disc, it gives the values of x ↦ U(a / x) - U(0) on the unit circle (CircleSobolev.toContinuousMap_neckTransfer). It is defined for ‖a‖ ≤ 1, where its norm is at most ‖a‖ (CircleSobolev.opNorm_neckTransfer_le); for ‖a‖ > 1 it is the junk value 0.

Instances For
    @[simp]
    theorem TauCeti.CircleSobolev.coeff_neckTransfer {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) (u : CircleSobolev s E) (n : ℤ) :
    ((neckTransfer a) u).coeff n = if n < 0 then a ^ n.natAbs • u.coeff (-n) else 0

    The Fourier coefficients of the neck transfer: the (-n)-th one, n > 0, is aⁿ uₙ, and the nonnegative ones vanish.

    @[simp]
    theorem TauCeti.CircleSobolev.coeff_neckTransfer_of_nonneg {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (u : CircleSobolev s E) {n : ℤ} (hn : 0 ≤ n) :
    ((neckTransfer a) u).coeff n = 0

    The neck transfer has no nonnegative frequencies.

    theorem TauCeti.CircleSobolev.coeff_neckTransfer_neg_natCast {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) (u : CircleSobolev s E) {n : ℕ} (hn : n ≠ 0) :
    ((neckTransfer a) u).coeff (-↑n) = a ^ n • u.coeff ↑n

    The neck transfer moves the n-th Fourier coefficient, n > 0, to frequency -n and multiplies it by aⁿ.

    @[simp]

    Outside the closed unit disc the neck transfer is the junk value 0.

    @[simp]

    The neck transfer of a node, a = 0, vanishes.

    The neck transfer has norm at most ‖a‖, for every a (with the junk value 0 when ‖a‖ > 1).

    The operator norm of the neck transfer is at most ‖a‖.

    The neck transfer has only negative frequencies.

    @[simp]
    theorem TauCeti.CircleSobolev.neckTransfer_indicator {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {S : Set ℤ} (hS : Set.Ioi 0 ⊆ S) (a : ℂ) (u : CircleSobolev s E) :

    The neck transfer only depends on the positive frequencies.

    The neck transfer vanishes on elements without positive frequencies.

    @[simp]
    theorem TauCeti.CircleSobolev.neckTransfer_single_natCast {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) {n : ℕ} (hn : n ≠ 0) (x : E) :
    (neckTransfer a) ((single ↑n) x) = (single (-↑n)) (a ^ n • x)

    The neck transfer of a Fourier monomial of positive frequency n is the monomial of frequency -n scaled by aⁿ.

    @[simp]
    theorem TauCeti.CircleSobolev.inclusion_neckTransfer {s t : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] (h : t ≤ s) (a : ℂ) (u : CircleSobolev s E) :

    The neck transfer commutes with the inclusion H^s ⊆ H^t.

    theorem TauCeti.CircleSobolev.toContinuousMap_neckTransfer {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} {T : ℝ} [Fact (0 < T)] [CompleteSpace E] (hs : 1 / 2 < s) (ha : ‖a‖ ≤ 1) (u : CircleSobolev s E) (x : AddCircle T) :
    ((toContinuousMap T hs) ((neckTransfer a) u)) x = u.diskFunction (a / (fourier 1) x) - u.coeff 0

    Boundary values of the neck transfer. For s > 1/2 and ‖a‖ ≤ 1, the continuous function on the circle attached to neckTransfer a u is x ↦ diskFunction u (a / x) - u₀, where x is read as a point fourier 1 x of the unit circle.

    Holomorphic dependence on the plumbing parameter #

    Holomorphy of the neck transfer. The neck transfer depends holomorphically on the plumbing parameter a in the open unit disc, as a map to the bounded operators on H^s.

    The derivative of the neck transfer at the node a = 0 is the operator u ↦ u₁ e^{-iθ}, which is nonzero for E ≠ 0 although the neck transfer itself vanishes there.

    The linearized node equation #

    The neck graph of the plumbing xy = a: the pairs (F, G) of elements of H^s whose Fourier coefficients satisfy F₋ₙ = -aⁿ Gₙ and G₋ₙ = -aⁿ Fₙ for every n > 0. This is the linearized node equation F(x) + G(a / x) = const on the neck, written in the Fourier coefficients on its two outer boundary circles (CircleSobolev.hasSum_zpow_smul_coeff_of_mem_neckGraph).

    Instances For
      theorem TauCeti.CircleSobolev.mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} {p : CircleSobolev s E × CircleSobolev s E} :
      p ∈ neckGraph s E a ↔ ∀ (n : ℕ), n ≠ 0 → p.1.coeff (-↑n) = -(a ^ n • p.2.coeff ↑n) ∧ p.2.coeff (-↑n) = -(a ^ n • p.1.coeff ↑n)

      Membership in the neck graph, in terms of the Fourier coefficients.

      The neck graph is closed in H^s × H^s.

      The neck graph is a graph. For ‖a‖ ≤ 1, a pair (F, G) lies in the neck graph exactly when the negative-frequency parts of F and G are -neckTransfer a G and -neckTransfer a F; these only depend on the positive frequencies of G and F.

      @[simp]

      At the node a = 0 the neck graph is the pair of nonnegative Laurent subspaces, the boundary values of holomorphic functions on the two branch discs.

      theorem TauCeti.CircleSobolev.sub_neckTransfer_mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) {F G : CircleSobolev s E} (hF : F ∈ supported s E (Set.Ici 0)) (hG : G ∈ supported s E (Set.Ici 0)) :

      Existence in the neck graph. Every pair (F, G) of nonnegative-frequency elements is the nonnegative part of the element (F - neckTransfer a G, G - neckTransfer a F) of the neck graph.

      theorem TauCeti.CircleSobolev.eq_of_mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) {p : CircleSobolev s E × CircleSobolev s E} (hp : p ∈ neckGraph s E a) :
      p = ((indicator (Set.Ici 0)) p.1 - (neckTransfer a) p.2, (indicator (Set.Ici 0)) p.2 - (neckTransfer a) p.1)

      Uniqueness in the neck graph. For ‖a‖ ≤ 1, an element of the neck graph is determined by its nonnegative-frequency parts.

      Transversality of the neck graph. For ‖a‖ ≤ 1 the neck graph is a complement of the pair of negative Laurent subspaces.

      The neck graph as a graph over the nonnegative Laurent subspaces #

      noncomputable def TauCeti.CircleSobolev.neckPoint {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] (a : ℂ) (PQ : ↥(supported s E (Set.Ici 0)) × ↥(supported s E (Set.Ici 0))) :

      The pair (P - neckTransfer a Q, Q - neckTransfer a P), whose nonnegative parts are (P, Q) (CircleSobolev.indicatorProj_sub_neckTransfer). For ‖a‖ ≤ 1 it lies in the neck graph (CircleSobolev.neckPoint_mem_neckGraph).

      Instances For
        @[simp]
        theorem TauCeti.CircleSobolev.neckPoint_fst {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] (a : ℂ) (PQ : ↥(supported s E (Set.Ici 0)) × ↥(supported s E (Set.Ici 0))) :
        (neckPoint a PQ).1 = ↑PQ.1 - (neckTransfer a) ↑PQ.2
        @[simp]
        theorem TauCeti.CircleSobolev.neckPoint_snd {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] (a : ℂ) (PQ : ↥(supported s E (Set.Ici 0)) × ↥(supported s E (Set.Ici 0))) :
        (neckPoint a PQ).2 = ↑PQ.2 - (neckTransfer a) ↑PQ.1
        theorem TauCeti.CircleSobolev.neckPoint_mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} (ha : ‖a‖ ≤ 1) (PQ : ↥(supported s E (Set.Ici 0)) × ↥(supported s E (Set.Ici 0))) :
        @[simp]

        The nonnegative part of P - neckTransfer a u is P.

        For ‖a‖ ≤ 1, every element of the neck graph is the neckPoint of its nonnegative parts.

        theorem TauCeti.CircleSobolev.analyticAt_neckPoint {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] [CompleteSpace E] {X : Type u_2} [NormedAddCommGroup X] [NormedSpace ℂ X] {g : X → ℂ} {h : X → ↥(supported s E (Set.Ici 0)) × ↥(supported s E (Set.Ici 0))} {x : X} (hg : AnalyticAt ℂ g x) (hgx : ‖g x‖ < 1) (hh : AnalyticAt ℂ h x) :
        AnalyticAt ℂ (fun (y : X) => neckPoint (g y) (h y)) x

        neckPoint a (P, Q) depends holomorphically on (a, P, Q) for a in the open unit disc.

        theorem TauCeti.CircleSobolev.toContinuousMap_eq_diskFunction_sub_of_mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} {T : ℝ} [Fact (0 < T)] [CompleteSpace E] (hs : 1 / 2 < s) (ha : ‖a‖ ≤ 1) {p : CircleSobolev s E × CircleSobolev s E} (hp : p ∈ neckGraph s E a) (x : AddCircle T) :
        ((toContinuousMap T hs) p.1) x = p.1.diskFunction ((fourier 1) x) - p.2.diskFunction (a / (fourier 1) x) + p.2.coeff 0

        Boundary values in the neck graph are differences of the two branch disc functions. This formula also makes sense at a = 0 and uses no Laurent evaluation at a zero coordinate.

        theorem TauCeti.CircleSobolev.hasSum_zpow_smul_coeff_of_mem_neckGraph {s : ℝ} {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℂ E] {a : ℂ} {T : ℝ} [Fact (0 < T)] [CompleteSpace E] (hs : 1 / 2 < s) (ha₀ : a ≠ 0) {p : CircleSobolev s E × CircleSobolev s E} (hp : p ∈ neckGraph s E a) (x : AddCircle T) :
        HasSum (fun (n : ℤ) => (a / (fourier 1) x) ^ n • p.2.coeff n) (p.1.coeff 0 + p.2.coeff 0 - ((toContinuousMap T hs) p.1) x)

        The linearized node equation. For s > 1/2, a ≠ 0 and (F, G) in the neck graph, the Laurent series G(a / x) = ∑ₙ (a / x)ⁿ Gₙ converges at every point x = fourier 1 x of the unit circle, and F(x) + G(a / x) = F₀ + G₀, where F(x) is the continuous function attached to F.