The Atkin–Lehner bar is the adjugate, conjugated by the Fricke matrix #
Two different matrices in this development are called "Atkin–Lehner". The Hecke-ring
anti-involution of TauCeti/NumberTheory/HeckeRing/GL2/Gamma0/AtkinLehner.lean conjugates a
transpose by the diagonal w = diag(1, N), and the Fricke matrix of
TauCeti/NumberTheory/ModularForms/Fricke/Matrix.lean is W = !![0, -1; N, 0]. Both of those
files record that the two matrices are different; neither says how they are related. This file
says how.
The relation is that conjugating the bar by W deletes the transpose and leaves the adjugate:
adjugate x = W · bar x · W⁻¹.
Nothing here is deep — in size two, adjugate is itself a conjugated transpose, adjugate g = J · gᵀ · J⁻¹ for J = !![0, -1; 1, 0], so bar and adjugate differ by conjugation by J · w⁻¹,
and that matrix is W up to a scalar, which conjugation does not see. What the identity buys is
a translation: the Petersson adjoint produces adjugateGL, the α ↦ (det α) · α⁻¹ involution of
TauCeti/NumberTheory/ModularForms/SlashAdjugate.lean, while the Hecke ring's commutativity
argument is phrased in the bar. The two layers can now be moved between.
Main results #
TauCeti.adjugateGL_eq_frickeGL_conj_bar:adjugate x = W · bar x · W⁻¹forx ∈ Δ₀(N).TauCeti.bar_eq_inv_frickeGL_conj_adjugateGL: the same read the other way,bar x = W⁻¹ · adjugate x · W.
References #
The Atkin–Lehner bar is the adjugate, conjugated by the Fricke matrix:
adjugate x = W · bar x · W⁻¹, where W = !![0, -1; N, 0].
The two "Atkin–Lehner" matrices of this development are the diagonal w = diag(1, N), which
conjugates the transpose to give the Hecke-ring bar, and W, the Fricke matrix. Since in size
two the adjugate is itself a conjugated transpose, the bar and the adjugate differ by a single
conjugation, and this identifies it.
The Atkin–Lehner bar, recovered from the adjugate: bar x = W⁻¹ · adjugate x · W. This
is adjugateGL_eq_frickeGL_conj_bar read in the other direction.