The twisted boundary of the Pin⁺ lift of a Kummer representation #
Let K be a field in which 2 ≠ 0, let L/K be a quadratic extension with an embedding
σ : L → Kˢ, let a ∈ Lˣ, and let ρ_a = Ind α_a : G_K → C₂ ≀ C₂ be the representation induced
from the Kummer character of a on G_L (TauCeti.kummerInd). Through the signed-permutation
representation C₂ ≀ C₂ ⊂ O₂, it is the action of G_K on the roots of X² − σ a and of its
conjugate. For a square root r2 of 2 in Kˢ, composing ρ_a with the lift
TauCeti.pinLift : C₂ ≀ C₂ → D₁₆ ⊂ M₂(Kˢ) gives a lift \tilde{ρ}_a of ρ_a to Pin⁺
(TauCeti.kummerIndLift).
The matrix \tilde{ρ}_a(g) has entries in Kˢ, on which G_K acts, so \tilde{ρ}_a has
a twisted boundary
δ(\tilde{ρ}_a)(g, h) = \tilde{ρ}_a(g) · g(\tilde{ρ}_a(h)) · \tilde{ρ}_a(g h)⁻¹
(TauCeti.twistedBoundary). Two signs enter it. The factor set of pinLift is the D₁₆
extension cocycle c_{D₁₆}; and g fixes e₁ and e₂ but sends t = (e₁ − e₂)/√2 to ±t
according to the sign rootSign r2 g of √2, so it multiplies pinLift w by
(−1)^{rootSign r2 g · c(w)}, where c(w) is the swap coordinate of w
(TauCeti.map_pinLift_galois). The swap coordinate of ρ_a(h) is the character of G_K with
kernel G_L, which is rootSign (σ x) h for any x ∈ L ∖ K
(TauCeti.coordC_kummerInd, in TauCeti.FieldTheory.GaloisCohomology.MuTwo.EvensNorm).
Hence, as an identity of matrices,
δ(\tilde{ρ}_a)(g, h) = (−1)^{c_{D₁₆}(ρ_a g, ρ_a h) + rootSign √2 g · rootSign x h}
(TauCeti.twistedBoundary_kummerIndLift): for x a square root of the discriminant d of
L/K, this is δ(\tilde{ρ}_a) = ρ_a^* c_{D₁₆} + (2) ∪ (d) at cochain level. The 𝔽₂-reading of
the twisted boundary (TauCeti.twistedBoundaryF2) records its exponent
(TauCeti.twistedBoundaryF2_kummerIndLift).
This is the cochain identity by which Serre's second proof of his Théorème 1′ compares the
Evens norm N^{Ev}((a)), the class of ρ_a^* c_{D₁₆}, with the second Stiefel–Whitney class of
the trace form Tr_*⟨a⟩: the correction (2) ∪ (d) involves the discriminant of L/K, and not
that of Tr_*⟨a⟩.
Main definitions #
TauCeti.twistedBoundary: the twisted boundary of aGL₂(Kˢ)-valued1-cochain ofG_K.TauCeti.twistedBoundaryF2: the twisted boundary read in𝔽₂.TauCeti.kummerIndLift: the lift\tilde{ρ}_a = pinLift ∘ ρ_a.
Main results #
TauCeti.twistedBoundary_conj: changing frame byQconjugates the twisted boundary byQ.TauCeti.twistedBoundary_pinDiagonalLift_rootSign: the boundary of a diagonal root-sign lift.TauCeti.map_pinLift_galois:g(pinLift w) = (−1)^{rootSign √2 g · c(w)} pinLift w.TauCeti.twistedBoundary_kummerIndLift,TauCeti.twistedBoundaryF2_kummerIndLift:δ(\tilde{ρ}_a)(g, h) = (−1)^{c_{D₁₆}(ρ_a g, ρ_a h) + rootSign √2 g · rootSign (σ x) h}for anyx ∈ L ∖ K; forxa square root of the discriminantd ∈ KofL/K, this isδ(\tilde{ρ}_a) = ρ_a^* c_{D₁₆} + (2) ∪ (d)at cochain level.
References #
- J.-P. Serre, L'invariant de Witt de la forme Tr(x²), Comment. Math. Helv. 59 (1984), 651–676, second proof of Théorème 1′.
- B. Kahn, Classes de Stiefel-Whitney de formes quadratiques et de représentations galoisiennes réelles, Invent. Math. 78 (1984), 223–256.
The Galois action on the lift #
The Galois action on the lift pinLift: g ∈ G_K fixes e₁ and e₂ and sends
t = (e₁ − e₂)/√2 to (−1)^{rootSign r2 g} t, since g r2 = ±r2; so it multiplies pinLift w by
(−1)^{rootSign r2 g · c(w)}, where c(w) is the top coordinate of w, which counts the
factors t.
Frame changes and sign twists #
The twisted boundary of a diagonal lift of Kummer characters is a cup product: if
c₀², c₁² ∈ K, the twisted boundary of g ↦ e₁^{rootSign c₀ g} e₂^{rootSign c₁ g} is
(g, h) ↦ (−1)^{rootSign c₁ g · rootSign c₀ h}, the sign of the cup product of the Kummer
characters of c₁² and c₀².
The lift of the Kummer representation #
The lift \tilde{ρ}_a = pinLift ∘ ρ_a of the representation
ρ_a = kummerInd σ hdeg a r hr s hs into D₁₆ ⊂ M₂(Kˢ), for a square root r2 of 2 in Kˢ.
Instances For
\tilde{ρ}_a(g) is the pinLift of ρ_a(g).
δ(\tilde{ρ}_a) = ρ_a^* c_{D₁₆} + (2) ∪ (d) at cochain level, as an identity of matrices:
δ(\tilde{ρ}_a)(g, h) = (−1)^{c_{D₁₆}(ρ_a g, ρ_a h) + rootSign √2 g · rootSign x h} for a generator
x of L = K(x), such as a square root of the discriminant d. The factor set of pinLift is
c_{D₁₆} (TauCeti.pinLift_mul_mul_inv), g twists pinLift w by the sign
(−1)^{rootSign √2 g · c(w)} (TauCeti.map_pinLift_galois), and the top coordinate of ρ_a h is
rootSign (σ x) h (TauCeti.coordC_kummerInd). This (2) ∪ (d) is Serre's (2)(d_E): it is
the discriminant of L/K that enters, and not that of Tr_*⟨a⟩.
δ(\tilde{ρ}_a) = ρ_a^* c_{D₁₆} + (2) ∪ (d) at cochain level, in 𝔽₂: the twisted boundary
of \tilde{ρ}_a, read in 𝔽₂, is c_{D₁₆}(ρ_a g, ρ_a h) + rootSign √2 g · rootSign x h.