A Lipschitz parametrization of the frontier of the ray fundamental domain #
TauCeti.NumberTheory.GeometryOfNumbers.LatticePointCount counts lattice points in a dilated
region with a power-saving error term, but only for regions whose frontier is Lipschitz
parametrizable in codimension one. NormLeOneLipschitz discharges that hypothesis for Mathlib's
normLeOne K, the norm-≤-one section of the fundamental cone. This file lifts it to the section
rayFundamentalDomain 𝔪 ∩ {x | mixedEmbedding.norm x ≤ 1} of the ray fundamental domain of an
arbitrary modulus, which is the region whose lattice points count the algebraic integers in a
fixed ray class.
rayFundamentalDomain_inter_normLeOne_eq presents that section as posRegion 𝔪 ∩ A, where
A = ⋃ q, rayUnitRepresentative 𝔪 q • normLeOne K is a finite union of unit translates: the
unit action preserves the mixed norm, so it commutes with the norm condition.
The two factors are of opposite character. A is bounded, with a complicated boundary;
posRegion 𝔪 is an unbounded finite intersection of open half spaces, with a boundary made of
hyperplanes. So the naive frontier (A ∩ B) ⊆ frontier A ∪ frontier B is useless here: the
frontier of posRegion 𝔪 is unbounded, and a Lipschitz-parametrizable set is a finite union of
Lipschitz images of a compact cube, hence bounded — so that union is parametrizable in no
dimension whatsoever. Mathlib's sharp form frontier_inter_subset keeps each frontier paired
with the closure of the other factor, and that pairing is what makes the argument work:
frontier A ∩ closure (posRegion 𝔪)lies infrontier A, which lies in the union of the frontiers of the finitely many translates; eachfrontier (u • normLeOne K)is a Lipschitz image offrontier (normLeOne K), because a unit acts by a homeomorphism;closure A ∩ frontier (posRegion 𝔪)is a bounded subset of finitely many coordinate hyperplanes, and a bounded subset of a hyperplane is Lipschitz parametrizable in codimension one (TauCeti.IsLipschitzParametrizable.of_isBounded_of_subset_ker).
Main results #
TauCeti.GlobalNumberFields.isLipschitzParametrizable_frontier_rayFundamentalDomain: the norm-≤-one section of the ray fundamental domain is bounded and measurable, and its frontier is Lipschitz parametrizable in dimensionfinrank ℝ (mixedSpace K) - 1, which is[K:ℚ] - 1bymixedEmbedding.finrank. These are exactly the three hypotheses the lattice-point count with a power-saving error consumes, so they are stated together;TauCeti.GlobalNumberFields.isBounded_rayFundamentalDomain_inter_normLeOneandTauCeti.GlobalNumberFields.measurableSet_rayFundamentalDomain_inter_normLeOne: the first two conclusions on their own, for callers that need only one of them;TauCeti.GlobalNumberFields.rayFundamentalDomain_inter_normLeOne_eq: that section is the positivity region cut by a finite union of unit translates ofnormLeOne K;TauCeti.GlobalNumberFields.frontier_posRegion_subset: the frontier of the positivity region lies in the coordinate hyperplanes prescribed by the infinite part of the modulus.
References #
- C. Birkbeck, AINTLIB at commit
db14b34cc5e3d79603e67c205dfa86b7b989000c(Apache-2.0),projects/Chebotarev/CebotarevDensity/ForMathlib/IdealCongruenceCount.lean, which carries out the same argument for a sign orthant cut out of a bounded region ofι → ℝ:frontier_posRegion_subsethere is that file'sfrontier_signOrthant_subset, andisLipschitzParametrizable_frontier_rayFundamentalDomainfollows itsexists_frontier_cover_inter_orthant, including the use offrontier_inter_subsetto pair each frontier with the other factor's closure. The bounded hyperplane pieces are handled here by the generalTauCeti.IsLipschitzParametrizable.of_isBounded_of_subset_kerrather than by that file's explicit slab chartexists_lipschitz_cube_cover_hyperplane_slab.
The frontier of the positivity region lies in the prescribed coordinate hyperplanes. The
region is cut out by finitely many strict inequalities, so a boundary point satisfies all of them
non-strictly yet must fail one of them: that coordinate vanishes. Mathlib's
frontier_lt_subset_eq is the single-inequality case; the content here is that a finite
intersection of such regions still confines its frontier to those level sets.
The norm-≤-one section of the ray fundamental domain, as the positivity region cut by a finite union of unit translates of Mathlib's norm-≤-one region. The unit action preserves the mixed norm, so it commutes with the norm condition.
The norm-≤-one section of the ray fundamental domain is bounded. It is carved out of a finite union of unit translates of Mathlib's norm-≤-one region, and each translate is bounded because the unit acts by a Lipschitz map.
The norm-≤-one section of the ray fundamental domain is measurable: the domain itself is measurable and the mixed norm is continuous.
The norm-≤-one section of the ray fundamental domain is bounded and measurable, and its
frontier is Lipschitz parametrizable in codimension one. The first and third conclusions are
exactly the two hypotheses hDb and hDfr of
TauCeti.exists_abs_ncard_smul_inter_vadd_sub_le, the lattice-point count with a power-saving
error, applied to the region counting the algebraic integers of a fixed ray class; the second is
what gives that region a measure at all. This is the analogue, for an arbitrary modulus, of
isLipschitzParametrizable_frontier_normLeOne for the trivial one, whose ray fundamental domain
is Mathlib's fundamental cone.
The section is posRegion 𝔪 intersected with finitely many unit translates of normLeOne K.
The translates are bounded and each has a frontier that is a Lipschitz image of
frontier (normLeOne K); the positivity region is unbounded, but frontier_inter_subset pairs
its frontier with the closure of the translates, so the piece it contributes is a bounded subset
of the finitely many coordinate hyperplanes prescribed by the infinite part of 𝔪.