The Fricke and Atkin–Lehner operators are Petersson-unitary #
The Fricke matrix W_N = !![0, -1; N, 0] normalises Γ₁(N), and an Atkin–Lehner matrix W_Q
for an exact divisor Q ∥ N normalises Γ₀(N). Their determinants are N, respectively Q,
so for N ≠ 1, respectively Q ≠ 1, they do not lie in SL₂(ℤ). Slashing both arguments of
the Petersson product by such a matrix of determinant D multiplies the product by D ^ (k - 2)
(TauCeti.CuspForm.peterssonInnerCosets_slash_of_inv_conjAct_eq), and the arithmetic
normalization 𝒲_Q = (√Q) ^ (2 - k) • (· ∣[k] W_Q) is exactly the one that cancels this factor:
the normalizer is real and its square is Q ^ (2 - k). So the normalized operators are
unitary for the Petersson product,
⟪𝒲_N f, 𝒲_N g⟫ = ⟪f, g⟫ on S_k(Γ₁(N)), ⟪𝒲_Q f, 𝒲_Q g⟫ = ⟪f, g⟫ on S_k(Γ₀(N)).
Combined with the square laws 𝒲_N² = (-1) ^ k and 𝒲_Q² = 1, unitarity gives the adjoints:
𝒲_Q is self-adjoint on S_k(Γ₀(N)), and 𝒲_N is self-adjoint up to the sign (-1) ^ k
on S_k(Γ₁(N)). Consequently the Petersson-orthogonal complement of a subspace stable under one
of these operators is again stable under it. That is the form in which the Fricke and
Atkin–Lehner operators reach the newforms: the new subspace is the Petersson-orthogonal complement
of the old subspace, so an operator preserving the old subspace preserves the new one, and on a
newform of trivial nebentypus multiplicity one then turns 𝒲_Q f into a multiple ± f — the
Atkin–Lehner sign.
Main results #
TauCeti.peterssonInnerCosets_frickeOperatorCusp: the raw Fricke operator scales the Petersson product byN ^ (k - 2).TauCeti.peterssonInnerCosets_normalizedFrickeOperatorCusp:𝒲_Nis unitary.TauCeti.peterssonInnerCosets_normalizedFrickeOperatorCusp_left: its adjoint is(-1) ^ k • 𝒲_N.TauCeti.normalizedFrickeOperatorCusp_mem_peterssonOrthogonal: the orthogonal complement of a𝒲_N-stable subspace is𝒲_N-stable.TauCeti.Nat.IsExactDivisor.peterssonInnerCosets_atkinLehnerOperatorCusp: the raw Atkin–Lehner operator scales the Petersson product byQ ^ (k - 2).TauCeti.Nat.IsExactDivisor.peterssonInnerCosets_normalizedAtkinLehnerOperatorCusp:𝒲_Qis unitary.TauCeti.Nat.IsExactDivisor.peterssonInnerCosets_normalizedAtkinLehnerOperatorCusp_left:𝒲_Qis self-adjoint.TauCeti.Nat.IsExactDivisor.normalizedAtkinLehnerOperatorCusp_mem_peterssonOrthogonal: the orthogonal complement of a𝒲_Q-stable subspace is𝒲_Q-stable.
References #
- F. Diamond and J. Shurman, A First Course in Modular Forms, §5.5 and §5.10.
- Miyake, Modular forms, Sections 4.5 and 4.6.
- A. O. L. Atkin and J. Lehner, Hecke operators on
Γ₀(m), Math. Ann. 185 (1970), 134–160.
The Fricke operator #
The raw Fricke operator scales the Petersson product by N ^ (k - 2):
⟪f ∣[k] W_N, g ∣[k] W_N⟫ = N ^ (k - 2) · ⟪f, g⟫ on S_k(Γ₁(N)), the determinant of
W_N = !![0, -1; N, 0] being N.
The normalized Fricke operator is Petersson-unitary: ⟪𝒲_N f, 𝒲_N g⟫ = ⟪f, g⟫ on
S_k(Γ₁(N)), in every weight. The normalizer (√N) ^ (2 - k) is real, so it contributes its
square N ^ (2 - k), cancelling the factor N ^ (k - 2) of the raw operator.
The Petersson adjoint of the normalized Fricke operator is (-1) ^ k • 𝒲_N:
⟪𝒲_N f, g⟫ = (-1) ^ k · ⟪f, 𝒲_N g⟫ on S_k(Γ₁(N)). Unitarity and the square law
𝒲_N² = (-1) ^ k combine to this; in even weight 𝒲_N is self-adjoint.
The Petersson-orthogonal complement of a 𝒲_N-stable subspace is 𝒲_N-stable. The
adjoint of 𝒲_N is a scalar multiple of 𝒲_N itself
(peterssonInnerCosets_normalizedFrickeOperatorCusp_left), so it preserves V whenever 𝒲_N
does.
The Atkin–Lehner operators #
The raw Atkin–Lehner operator scales the Petersson product by Q ^ (k - 2):
⟪f ∣[k] W_Q, g ∣[k] W_Q⟫ = Q ^ (k - 2) · ⟪f, g⟫ on S_k(Γ₀(N)), the determinant of an
Atkin–Lehner matrix for Q being Q.
The normalized Atkin–Lehner operator is Petersson-unitary: ⟪𝒲_Q f, 𝒲_Q g⟫ = ⟪f, g⟫ on
S_k(Γ₀(N)). The real normalizer (√Q) ^ (2 - k) cancels the factor Q ^ (k - 2) of the raw
operator.
The normalized Atkin–Lehner operator is Petersson self-adjoint:
⟪𝒲_Q f, g⟫ = ⟪f, 𝒲_Q g⟫ on S_k(Γ₀(N)), a unitary involution being its own adjoint.
The Petersson-orthogonal complement of a 𝒲_Q-stable subspace is 𝒲_Q-stable, 𝒲_Q
being self-adjoint.