Documentation

LeanPool.BrillNoetherGraphs.Utilities.Subdivision.OddSubdivisionDescent

Odd subdivision descent for Brill--Noether rank #

Utilities/Subdivision/SubdivisionChipDescent.lean rounds finitely many moving chips on an N-fold refinement to chosen ends of their coarse steps, provided the total rounding distance is less than N. This file chooses the ends: every fine vertex is rounded to its nearest coarse vertex, at distance at most N / 2, and strictly less than N / 2 when N is odd. Two chips on an odd refinement therefore always fit the budget.

The consequence for the discrete Brill--Noether rank w^r_d of Utilities/Foundations/BrillNoetherRank.lean is:

None of this needs a genus hypothesis or any algebraic-geometric input. The production of an odd refinement carrying w^1_4 ≥ 1 (or of the pairwise witnesses) for a genus-five graph is a separate matter, discussed in the private research notes; this file proves only the descent.

Rounding every fine vertex to its nearest coarse vertex #

def Utilities.Certificate.SubdivisionGraph.Spec.roundData {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) :
(spec.scale N hN).Vertex → spec.Vertex ⊕ spec.Chip N

Classify a fine vertex: either it is the image of a coarse vertex, or it is a chip strictly inside a coarse step, rounded to the nearer end (ties go left).

Equations
Instances For
    def Utilities.Certificate.SubdivisionGraph.Spec.nearest {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (y : (spec.scale N hN).Vertex) :
    spec.Vertex

    The nearest coarse vertex of a fine vertex (ties rounded to the left).

    Equations
    Instances For
      def Utilities.Certificate.SubdivisionGraph.Spec.roundDist {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (y : (spec.scale N hN).Vertex) :

      The fine distance from a fine vertex to its nearest coarse vertex.

      Equations
      Instances For
        theorem Utilities.Certificate.SubdivisionGraph.Spec.eq_fineOf_of_roundData_eq_inl {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) {y : (spec.scale N hN).Vertex} {x : spec.Vertex} (h : spec.roundData N hN y = Sum.inl x) :
        y = spec.fineOf N hN x
        theorem Utilities.Certificate.SubdivisionGraph.Spec.eq_fineVertex_of_roundData_eq_inr {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) {y : (spec.scale N hN).Vertex} {c : spec.Chip N} (h : spec.roundData N hN y = Sum.inr c) :
        theorem Utilities.Certificate.SubdivisionGraph.Spec.two_mul_distance_le_of_roundData_eq_inr {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) {y : (spec.scale N hN).Vertex} {c : spec.Chip N} (h : spec.roundData N hN y = Sum.inr c) :
        2 * c.distance ≤ N
        theorem Utilities.Certificate.SubdivisionGraph.Spec.two_mul_roundDist_le {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (y : (spec.scale N hN).Vertex) :
        2 * spec.roundDist N hN y ≤ N
        theorem Utilities.Certificate.SubdivisionGraph.Spec.two_mul_roundDist_lt {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (hodd : Odd N) (y : (spec.scale N hN).Vertex) :
        2 * spec.roundDist N hN y < N
        theorem Utilities.Certificate.SubdivisionGraph.Spec.embed_one_chip {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (x : spec.Vertex) :
        spec.embed N hN (oneChip x) = oneChip (spec.fineOf N hN x)
        theorem Utilities.Certificate.SubdivisionGraph.Spec.rank_ge_of_rank_scale_ge_nearest {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) {ι : Type u_1} [Fintype ι] (ys : ι → (spec.scale N hN).Vertex) (D₀ : CFDiv spec.graph) (r : ℤ) (hbudget : ∑ i : ι, ↑(spec.roundDist N hN (ys i)) < ↑N) (hrank : rank (spec.scale N hN).graph (spec.embed N hN D₀ + ∑ i : ι, oneChip (ys i)) ≥ r) :
        rank spec.graph (D₀ + ∑ i : ι, oneChip (spec.nearest N hN (ys i))) ≥ r

        Descent of rank with nearest rounding. Any family of fine chips whose total nearest-rounding distance is less than N may be rounded to nearest coarse vertices without lowering any rank bound of the embedded divisor plus the chips.

        Brill--Noether rank descends along odd refinements #

        theorem Utilities.Certificate.SubdivisionGraph.Spec.bnRankGe_of_bnRankGe_scale_two {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) (hodd : Odd N) {r d k : ℤ} (hd : d = r + k + 2) (h : BNRankGe (spec.scale N hN).graph r d k) :
        BNRankGe spec.graph r d k

        Two residual chips, odd scale. If d = r + k + 2 and N is odd, then w^r_d(σ_N) ≥ k implies w^r_d ≥ k on the coarse graph.

        theorem Utilities.Certificate.SubdivisionGraph.Spec.bnRankGe_of_bnRankGe_scale_one {n p : ℕ} (spec : Spec n p) (N : ℕ) (hN : 0 < N) {r d k : ℤ} (hd : d = r + k + 1) (h : BNRankGe (spec.scale N hN).graph r d k) :
        BNRankGe spec.graph r d k

        One residual chip, any scale. If d = r + k + 1, then w^r_d(σ_N) ≥ k implies w^r_d ≥ k on the coarse graph for every N ≥ 1.

        noncomputable def Utilities.Gonality.regularSubdivisionVertex (G : CFGraph) (N : ℕ) (hN : 0 < N) (v : G.V) :

        The image of an original vertex of G in its regular subdivision σ_N G.

        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          theorem Utilities.Gonality.bnRankGe_of_bnRankGe_regularSubdivision (G : CFGraph) {N : ℕ} (hN : 0 < N) (hodd : Odd N) {r d k : ℤ} (hd : d = r + k + 2) (h : BNRankGe (regularSubdivision G N hN) r d k) :
          BNRankGe G r d k

          Odd subdivision descent for Brill--Noether rank, two residual chips. For every finite loopless multigraph G, odd N, and d = r + k + 2, w^r_d(σ_N G) ≥ k implies w^r_d(G) ≥ k.

          The genus-five descent statement. If some odd regular subdivision of G has w^1_4 ≥ 1, so does G itself. No genus hypothesis is needed.

          The odd pair witness. Every pair of original vertices x, y (equal or not) admits, on some odd regular subdivision depending on the pair, an effective degree-two completion of x + y to a divisor of rank at least one. This is the exact combinatorial input that the algebraic degree-five argument of the research notes is meant to produce for genus-five graphs.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For

            The pairwise form of the descent. An odd pair witness already gives w^1_4(G) ≥ 1; the odd scale may depend on the pair, and only the pencil through that pair is required on the refinement.