The order filtration of a function field at a place #
A place P of F / k filters F by the order of vanishing at P: for an integer a, zero
together with the nonzero functions satisfying ord_P z ≥ a forms the k-subspace
𝔪_P^a = {z ∈ F | v_P z ≤ exp (-a)}
of F, decreasing in a. At a = 0 its membership condition is v_P z ≤ 1, that of the
valuation ring 𝒪_P, and at a = 1 it is v_P z < 1, that of the maximal ideal 𝔪_P of 𝒪_P.
For negative a it is the fractional ideal of functions with a pole of order at most -a, so the
whole filtration lives inside F and no completion is taken.
This file constructs that filtration and computes its successive quotients. Multiplication by a
function of order -a identifies 𝔪_P^a / 𝔪_P^(a + 1) with the residue field F_P
(TauCeti.Place.filtrationQuotientEquivResidueField), and iterating this along the tower gives
the dimension formula
dim_k (𝔪_P^a / 𝔪_P^b) = (b - a) · deg P for a ≤ b
(TauCeti.Place.finrank_quotient_filtration). This is the local half of the local-to-global
engine of Stichtenoth's Section I.5: the global half identifies the quotient A_F(E) / A_F(D)
of two members of the divisor filtration of the repartition space with a finite direct sum of
these local quotients, and so computes dim_k (A_F(E) / A_F(D)) = deg E - deg D.
Main definitions #
TauCeti.Place.filtration: the subspace𝔪_P^a ⊆ F.TauCeti.Place.filtrationResidue: forsof order-a, thek-linear evaluation𝔪_P^a → F_P,z ↦ (s · z)(P).TauCeti.Place.filtrationQuotientEquivResidueField: the induced isomorphism𝔪_P^a / 𝔪_P^(a + 1) ≃ₗ[k] F_P.
Main results #
TauCeti.Place.rank_quotient_filtration_add_one: one step of the filtration has the rank of the residue field.TauCeti.Place.mem_maximalIdeal_pow_iff_coe_mem_filtration: the positive part of the place filtration is the maximal-ideal filtration of the valuation ring.TauCeti.Place.finrank_quotient_filtration_addandTauCeti.Place.finrank_quotient_filtration:dim_k (𝔪_P^a / 𝔪_P^b) = (b - a) · deg P, in the form indexed byb = a + nwithn : ℕand in the integer form.TauCeti.Place.finiteDimensional_quotient_filtration: those quotients are finite-dimensional as soon as the residue field is, which for a function field isTauCeti.Place.finiteDimensional_residueField.
Implementation notes #
Membership is stated multiplicatively, as v_P z ≤ exp (-a), and not in the unguarded additive form
ord_P z ≥ a, which the junk value ord_P 0 = 0 would get wrong at a > 0: the additive
carrier would not contain 0 at positive a and so would not be a subspace. This is the
convention of TauCeti.riemannRochSpace and TauCeti.adeleFiltration, of which this filtration
is the one-place shadow; TauCeti.Place.mem_filtration_iff_le_ord is the additive form, guarded
by z ≠ 0.
The relative quotients are spelled ↥(P.filtration a) ⧸ (P.filtration b).submoduleOf (P.filtration a), using Mathlib's Submodule.submoduleOf for the trace of the smaller subspace
on the larger one. The dimension formula is proved first as a statement about
Module.rank, where it needs no finiteness hypothesis at all, and then transported to
Module.finrank through Cardinal.toNat_mul; it therefore holds unconditionally, both sides
being zero when the residue field is infinite-dimensional over k.
References #
- H. Stichtenoth, Algebraic Function Fields and Codes, 2nd ed., GTM 254, Springer, 2009, Sections I.1 and I.5.
The filtration #
The subspace 𝔪_P^a whose nonzero elements are the functions satisfying ord_P z ≥ a
at P, for a an arbitrary integer: for a ≤ 0 it is the space of functions with a pole of
order at most -a at P, and for a > 0 the a-th power of the maximal ideal of 𝒪_P.
The defining condition is the multiplicative v_P z ≤ exp (-a), which is junk-free at z = 0;
TauCeti.Place.mem_filtration_iff_le_ord is the additive form.
Equations
- P.filtration a = Submodule.restrictScalars k (P.valuation.leSubmodule (WithZero.exp (-a)))
Instances For
The additive form of membership in 𝔪_P^a. The nonvanishing hypothesis guards the junk
value ord_P 0 = 0: the zero function lies in 𝔪_P^a for every a, whereas the inequality
a ≤ ord_P 0 fails precisely when a > 0.
Membership in 𝔪_P^1 is the maximal-ideal condition v_P z < 1
(TauCeti.Place.mem_maximalIdeal_iff_valuation_lt_one): 𝔪_P^1 is the maximal ideal of 𝒪_P,
seen inside F.
The positive part of the order filtration is the maximal-ideal filtration of the valuation
ring: an integral function belongs to 𝔪_P ^ n exactly when its image in the function field lies
in P.filtration n, that is, when v_P(x) ≤ exp (-n).
Two functions integral at P have the same value at P exactly when they differ by a
function of positive order: 𝔪_P^1 is the maximal ideal of 𝒪_P, seen inside F.
A function of order at least a and one of order at least b have a product of order at
least a + b.
A function integral at P and congruent to 1 to order a stays congruent to 1 to
order a after being raised to a power.
The successive quotients #
Multiplying a function of order at least a by one of order -a lands in the valuation
ring: the integrality behind the evaluation map TauCeti.Place.filtrationResidue.
Evaluation of s · z at P, for a fixed function s of order -a: a k-linear map from
𝔪_P^a to the residue field F_P, whose kernel is 𝔪_P^(a + 1) and which is surjective. It
is the local form of the evaluation map in Stichtenoth's proof of Lemma 1.4.8.
Defining the map needs nothing beyond hs; it is only its kernel and its surjectivity
(TauCeti.Place.ker_filtrationResidue, TauCeti.Place.filtrationResidue_surjective) that fail
at s = 0, so those carry the nonvanishing hypothesis.
Equations
- TauCeti.Place.filtrationResidue hs = { toFun := fun (z : ↥(P.filtration a)) => (IsScalarTower.toAlgHom k (↥P.integers) P.ResidueField) ⟨s * ↑z, ⋯⟩, map_add' := ⋯, map_smul' := ⋯ }
Instances For
The local quotient of the order filtration. Multiplication by a function s of order
-a, followed by evaluation at P, identifies 𝔪_P^a / 𝔪_P^(a + 1) with the residue field
F_P as k-vector spaces.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The quotient equivalence evaluates the residue of s * z on a representative z.
One step of the order filtration has the rank of the residue field.
The dimension of a local quotient of the order filtration, in the form indexed by a
natural number of steps: dim_k (𝔪_P^a / 𝔪_P^(a + n)) = n · deg P, at the level of ranks and so
without any finiteness hypothesis on the residue field.
The dimension of a local quotient of the order filtration (Stichtenoth, Section I.5): the
k-dimension of 𝔪_P^a / 𝔪_P^(a + n) is n · deg P.
Both sides are junk — zero — when the residue field is infinite-dimensional over k, so no
finiteness hypothesis is needed; for a function field the residue field is always
finite-dimensional (TauCeti.Place.finiteDimensional_residueField).
The dimension of a local quotient of the order filtration, in integer form: for a ≤ b,
dim_k (𝔪_P^a / 𝔪_P^b) = (b - a) · deg P. This is the local input of the local-to-global
engine computing dim_k (A_F(E) / A_F(D)) = deg E - deg D for the divisor filtration of the
repartition space.
The local quotients of the order filtration are finite-dimensional as soon as the residue
field is, which for a function field is TauCeti.Place.finiteDimensional_residueField. No
order relation between a and b is needed: for b ≤ a the quotient is trivial.