The discriminant and the invariant factors of ℤ[conj α] (biquadratic case) #
In the model V4Data O of BiquadraticModel, A = ℤ[x₀, …, x₃] is the subring generated by the
conjugates (Aring), ℤ[α] = ℤ + ℤα + ℤα² + ℤα³, indices are AddSubgroup.relIndex, and
D(ω) = det[γ_i(ω_l)]. The triangular basis of A (Louboutin 2020, Prop. 10; Cohen 1993,
Cor. 4.7.6) enters as the structure Louboutin.Triang for the image of A in ℚ⁴; it is used
only for d₁, d₂, d₃ and the isomorphism. d₁ = 1 is proved, not assumed.
The route: [A : L] = 2 for L = ⟨1, η₁, η₂, η₃⟩ is shown by exhibiting A = L + ℤy through
ring closure, where each product has an odd multiple in L + ℤy (a 2-adic computation) and a
power-of-2 multiple in ⟨1, ϑ_j⟩, glued by Bézout; indices are computed by determinants
(AddSubgroup.relIndex_eq_natAbs_det); d₂ is read from coordinates in the ϑ-basis, a global
replacement of a local Nakayama argument.
Main results #
Louboutin.DiscV4.thm_disc: the discriminant theorem, with(n_j, t_j),h₀,m₀,g.Louboutin.DiscV4.Dat.prop_discA:g ∣ Δ,[A : ℤ[α]] = |Δ|/g,g² = Δ₁Δ₂Δ₃, andD(b)² = g²for every basisbofA(sodisc A = g²).Louboutin.DiscV4.Dat.prop_smith:h₀ ∈ {1, 2},h₀ ∣ m₀,d₁ = 1,d₂ = m₀/h₀,d₃ = I h₀/m₀,A/ℤ[α] ≅ ℤ/(m₀/h₀) × ℤ/(I h₀/m₀), and{1, α, α², ω}is a basis iffm₀ = h₀.
References #
- S. R. Louboutin, On the order generated by the conjugates of an algebraic integer, in: Algebra, Logic and Number Theory, Banach Center Publ. 121 (2020), 107–124, doi:10.4064/bc121-10.
- S. R. Louboutin, When is the order generated by a cubic, quartic or quintic algebraic unit Galois invariant: three conjectures, Czechoslovak Math. J. 70 (2020), 905–919, doi:10.21136/CMJ.2020.0019-19.
- C. Marín, Orders generated by conjugates of Galois quartic integers, preprint (2026).
Part 1. Pure integer core #
Integer data extracted from the model (see Dat.toZD).
Instances For
Parity helpers #
Boolean parity core of the all-even case (2⁷ cases; A = a₀, M_j = m_j, E_j = e_j even).
Part 2. The V4 model: the note's pairs (n_j, t_j) and the elements ϑ_j #
With r² = D and rq = E (integers), the note's coprime pair (n, t) and some ϑ ∈ O satisfy
r = nϑ, q = tϑ (proof of prop:smith, first paragraph; cf. split_pair in Conj61Model).
A coprime splitting inherits the sign of the pair: τ r = e r, τ q = e q ⇒ τ ϑ = e ϑ.
The data of the note's proof of prop:discA / prop:smith in the model: the integers a, b, c, d, z_j, Δ = δ, the coprime pairs (n_j, t_j) with Bézout coefficients (u_j, v_j), the elements ϑ_j with r_j = n_j ϑ_j, q_j = t_j ϑ_j, Δ_j = ϑ_j², Θ = ϑ₁ϑ₂ϑ₃.
- a : ℤ
- b : ℤ
- c : ℤ
- d : ℤ
- z₁ : ℤ
- z₂ : ℤ
- z₃ : ℤ
- δ : ℤ
- n₁ : ℤ
- n₂ : ℤ
- n₃ : ℤ
- t₁ : ℤ
- t₂ : ℤ
- t₃ : ℤ
- u₁ : ℤ
- u₂ : ℤ
- u₃ : ℤ
- v₁ : ℤ
- v₂ : ℤ
- v₃ : ℤ
- ϑ₁ : O
- ϑ₂ : O
- ϑ₃ : O
- Δ₁ : ℤ
- Δ₂ : ℤ
- Δ₃ : ℤ
- Θ : ℤ
Instances For
Existence of the data with the note's (n_j, t_j) = ntN/ntT (D_j, E_j), D_j = a² − 4b + 4z_j, E_j = a z_j − 2c.
Part 3. Identities in the model #
η_j = u_j S_j + v_j Q_j with S_j = α + τ_jα, Q_j = α·τ_jα: the generator of
C_j = ℤ + ℤS_j + ℤQ_j (Lemma lem:Lambda), with ϑ_j = 2η_j − e_j.
Instances For
ϑ_j ≠ 0: (α − x_k)(α − x_l) = α r_j − q_j = (α n_j − t_j) ϑ_j.
2α − Σ n_j η_j is an integer s with a − Σ n_j e_j = 2s; if all n_j are even, 4 ∣ a − Σ n_j e_j.
Θ·d₁₂d₂₃d₃₁ = Δ·n₁n₂n₃ (from cross_prod; proof of cor:weak).
The integer data of the model.
Equations
- One or more equations did not get rendered due to their size.
Instances For
D(ω) = det[γ_i(ω_l)], the note's detG (γ₀ = id, γ_j = τ_j).
Instances For
Lemma lem:sub (general form): if ω'_l = Σ_k C_lk ω_k with C integral, D(ω') = D(ω) det C.
D(ω) ≠ 0 ⇒ ω₁, …, ω₄ are ℤ-independent.
The lattice spanned by a family.
Equations
Instances For
Index of lattices by determinants (AddSubgroup.relIndex_eq_natAbs_det):
[⟨v⟩ : ⟨w⟩] = |det C| where w = C v, and D(w) = D(v) det C.
Part 5. The order A = ℤ[x₁, …, x₄], Cramer's rule, and the closure arguments #
A = ℤ[conj(α)] = ℤ[x₁, x₂, x₃, x₄], the subring generated by the conjugates.
Equations
Instances For
The elements on which the γ's compose as the group V₄ (γ_a γ_i = γ_{a xor i}).
Equations
- One or more equations did not get rendered due to their size.
Instances For
the permutation k ↦ a xor k.
Equations
- Louboutin.DiscV4.xtPerm a = { toFun := Louboutin.Conj61.xt a, invFun := Louboutin.Conj61.xt a, left_inv := ⋯, right_inv := ⋯ }
Instances For
Cramer's rule with integer numerators (Lemma lem:cramer, first part, for every y ∈ A):
Δ·y ∈ ℤ[α].
Stability can be checked on generators.
Part 6. The lattices M = ⟨1, ϑ_j⟩ ⊆ L = ⟨1, η_j⟩ ⊆ A, and ℤ[α] #
the action of V₄ on the ϑ_j (each is an eigenvector of a nontrivial character).
Equations
- Louboutin.DiscV4.Dat.vpow M k = M.x 0 ^ ↑k
Instances For
Part 7. M[1/2] is a ring containing A; the indices of M, L, X, ℤ[α] #
4ϑ_j = (4u_j + a v_j) n_j ϑ_j − v_j n_k n_l ϑ_k ϑ_l (from 4q_j = a r_j − r_k r_l).
M = ℤ + ℤϑ₁ + ℤϑ₂ + ℤϑ₃ (as a ℤ-submodule of O).
Instances For
L = ℤ + ℤη₁ + ℤη₂ + ℤη₃.
Instances For
The indices #
[A : L] = 2 when A = L + ℤy with 2y ∈ L, y ∉ L.
prop:discA, the index: from [A : L] = 2, [A : ℤ[α]] · |Θ| = |Δ|.
Part 8. [A : L] = 2 (proof of prop:discA) #
The closure scheme: if every product x_m·g (g a generator of N = L + ℤy) has an odd multiple
in N, then A ⊆ N (the 2-power multiple is supplied by half_loc; Bézout glues).
Case I: some n_i odd, A = L + ℤα #
From the four triple products to all products x_m·g, g ∈ {1, η_j, α} (Case I).
Case II: all n_j even, A = L + ℤ η_i η_j #
Products in L·N for the configuration (i, j, k) with c η_k = 2m_im_j η_iη_j − … (an odd multiple of every product lands in N = L + ℤ η_iη_j). Abstract over a commutative ring.
η_iη_j ∉ ℤ + ℤη_i + ℤη_j + ℤη_k when c is odd (comparing η_k-coordinates).
From m_i m_j ϑ_i ϑ_j = c ϑ_k and ϑ = 2η − e: c η_k = 2m_im_j η_iη_j − m_im_j e_j η_i − m_im_j e_i η_j + K with K ∈ ℤ (proof of prop:discA, all-even case).
Case II of prop:discA for one configuration (i, j, k): A = L + ℤ η_iη_j, 2η_iη_j ∈ L, η_iη_j ∉ L.
[A : L] = 2 in all cases (proof of prop:discA).
gcd(T, U, V, W) = |Θ| for the note's (n_j, t_j) (Lemma lem:split, cf. gcd_TUVW).
Part 9. Integer core of prop:smith (the invariant factors) #
d₁ = 1, Case II: d(2c_i + c e_j) = 2m_i, d(2c_j + c e_i) = 2m_j with e_i + e_j odd, m_i, m_j odd and gcd(m_i, m_j) = 1 force d to be a unit.
d₂ ∣ m₀, Case I: (P − dc) n_l − 2t_l = 2d c_l for all l, with some n_i odd.
d₂ ∣ m₀/2, Case II (configuration (i, j, k): m_i, m_j odd, m_k even).
Part 10. Coordinates: A ↪ ℚ⁴ in the basis 1, α, α², α³ (bridge to T1Completo) #
the ℤ-basis 1, α, α², α³ of ℤ[α].
Equations
Instances For
integer coordinates of Δ·y.
Instances For
φ(y) = (coordinates of y in 1, α, α², α³) ∈ ℚ⁴.
Equations
- P.φ = { toFun := fun (y : ↥(Louboutin.DiscV4.Aring M).toAddSubgroup) (k : Fin 4) => ↑(P.rc y k) / ↑P.δ, map_zero' := ⋯, map_add' := ⋯ }
Instances For
d·y = Σ c_k αᵏ ⟺ φ(y) = c/d (d ≠ 0).
the image of A in ℚ⁴.
Instances For
[φ(A) : ℤ⁴ ∩ φ(A)] = [A : ℤ[α]].
Part 11. Coordinates on the ϑ-basis (comparison of coefficients) #
Case I, d₁: d·w = p + α with w = c₀ + Σ c_l η_l + κα gives (1 − dκ) n_l = 2d c_l.
Case I, d₂: d·w = q + pα + α² gives (a + p) n_l − 2t_l = 2d c_l + dκ n_l.
Case II, d₁ (configuration (i, j, k)): d·w = p + α with w = c₀ + Σ c_l η_l + κ η_iη_j.
Case II, d₂ (configuration (i, j, k)): d·w = q + pα + α².
Part 12. prop:smith in the model: d₁ = 1 and d₂ = m₀/h₀ for the triangular basis #
The element of A behind the lower bound: 2N α² = K₀ + K₁α + 2n₂d₁₃ x₁ + 2n₁d₂₃ x₂.
Case II of prop:smith for one configuration (i, j, k): d₁ = 1 and d₂ divides the three reduced minors m_i t_j − m_j t_i, … (= d_ij/2).
Part 13. Statements: prop:discA, prop:smith, thm:disc #
A ≃ φ(A) (additive), carrying ℤ[α] onto ℤ⁴.
Equations
Instances For
prop:discA (in the model): with g = gcd(T, U, V, W), g ∣ Δ_α, [A : ℤ[α]] = |Δ_α|/g, g² = Δ₁Δ₂Δ₃, and D(b)² = g² for every family b spanning A (disc A = D(b)², b a ℤ-basis of A).
prop:smith (in the model), for the triangular basis B of A given by prop:triang (cited): h₀ ∈ {1, 2}, h₀ ∣ m₀, d₁ = 1, d₂ = m₀/h₀, d₃ = I h₀/m₀ with I = [A : ℤ[α]], A/ℤ[α] ≅ ℤ/(m₀/h₀) × ℤ/(I h₀/m₀), and A has a ℤ-basis {1, α, α², ω} iff m₀ = h₀.
From A = L + ℤy with 2y = λ₀ + (2q+1)g₁ + λ₂g₂ + λ₃g₃ ({g₁, g₂, g₃} = {η₁, η₂, η₃}): the four elements 1, y − q g₁, g₂, g₃ span A.
A has a family of four elements spanning it (hence, by prop_discA, a ℤ-basis with
D(b)² = g² ≠ 0): the statement disc A = g² is not vacuous.
Part 14. Theorem B (thm:disc) in the V4 model #
the note's n_j = D_j/gcd(D_j, E_j) (or 0), D_j = a² − 4b + 4z_j, E_j = a z_j − 2c.
Equations
Instances For
the note's t_j = E_j/gcd(D_j, E_j) (or 1).
Equations
Instances For
the note's h₀ = gcd(n₁, n₂, n₃).
Equations
- Louboutin.DiscV4.h0N a b c z₁ z₂ z₃ = (↑((Louboutin.DiscV4.noteN a b c z₁).gcd (Louboutin.DiscV4.noteN a b c z₂))).gcd (Louboutin.DiscV4.noteN a b c z₃)
Instances For
the note's m₀ = gcd(d₁₂, d₁₃, d₂₃), d_ij = n_i t_j − n_j t_i.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Theorem B (thm:disc) in the abstract V4 model. Let a, b, c, d be the coefficients of
Π_α, z₁, z₂, z₃ the roots of the resolvent, Δ = Δ_α, T, U, V, W as in the note,
g = gcd(T, U, V, W), and (n_j, t_j), h₀, m₀ as in the note (noteN, noteT, h0N, m0N).
Then, for A = ℤ[x₁, …, x₄] and ℤ[α] = ℤ + ℤα + ℤα² + ℤα³ (Lat (Dat.vpow M)):
h₀ ∈ {1, 2}, h₀ ∣ m₀, g ∣ Δ, [A : ℤ[α]] = |Δ|/g, disc A = g² (D(b)² = g² for every family b
spanning A, and such b exist); and for the triangular basis of A (prop:triang, cited,
Triang, in the coordinates φ : A ↪ ℚ⁴ of Dat.MQ): d₁ = 1, d₂ = m₀/h₀,
d₃ = (|Δ|/g)h₀/m₀, A/ℤ[α] ≅ ℤ/(m₀/h₀) × ℤ/((|Δ|/g)h₀/m₀), and A has a ℤ-basis
{1, α, α², ω} iff m₀ = h₀.