The elliptic trace fibres of Popa–Zagier's explicit element #
Let TauCeti.PopaZagier.weight M be twelve times the coefficient of the integral matrix M in
Popa and Zagier's explicit Hecke element, their eq. (15). If t² - 4 n = -D < 0, then
∑_{tr M = t, det M = n} weight M = -12 H(D),
the sum running over all integral matrices of trace t and determinant n, where H is the
Hurwitz class number TauCeti.hurwitzClassNumber. In Popa and Zagier's normalisation: the
coefficients of their element sum to -H(4 n - t²) over every elliptic trace fibre. The trace of
a matrix on the binary forms of degree w only depends on its trace and determinant
(TauCeti.trace_binaryFormRep_eq_dickson_eval), so this identity is the input that evaluates the
contribution of the elliptic matrices to the trace of the element on binary forms, the
class-number terms P_{w+2}(t, n) H(4 n - t²) of the Eichler–Selberg trace formula.
Main results #
TauCeti.PopaZagier.finsum_weight_traceDetFiber_add_two: forn > 0, the weights of the matrices of tracet + 2and determinantn + t + 1have the same sum as those of tracetand determinantn.TauCeti.PopaZagier.finsum_weight_traceDetFiber_eq_neg_twelve_mul_hurwitzClassNumber: the weights of the matrices of tracetand determinantnsum to-12 H(D)whent² - 4 n = -D < 0.
Proof #
The translation M ↦ M + 1 maps the matrices of trace t and determinant n bijectively to
those of trace t + 2 and determinant n + t + 1. It fixes the form
Q_M = c x² + (d - a) x y - b y² of M = (a b; c d), hence the discriminant t² - 4 n, and it
fixes the first line of each of the four sums of eq. (15). So it changes the weight of M only
where one of the second-line inequalities c < a, b < d ≤ 0, a ≤ 0 and d ≤ 0 changes, that is
at the matrices with a = c, b = d, d = 0 or a = 0. For n > 0 these are parametrised by the
factorisations n = u v into positive integers, and their changes cancel under (u, v) ↦ (v, u).
This gives the first result; the sum over the matrices of trace t and determinant n therefore
depends only on D = 4 n - t².
For t ≤ -D only the sums T₃ and T₄ contribute, with all their second-line inequalities
automatic. Through M ↦ Q_M their matrices are the forms ⟨a', b', c'⟩ of discriminant -D with
0 ≤ -b' ≤ a' ≤ c', respectively 0 ≤ -b' ≤ c' ≤ a'. Each of the two families is, with its
weights, six times the weighted count of the reduced forms of discriminant -D that defines
H(D).
References #
- A. Popa and D. Zagier, An elementary proof of the Eichler–Selberg trace formula, J. Reine Angew. Math. 762 (2020), 105–122, arXiv:1711.00327, §1 and §4, eq. (15).
Translation by the identity #
Factorisations of n #
Translation by the identity. For n > 0, the weights of the matrices of trace t + 2
and determinant n + t + 1 have the same sum as those of trace t and determinant n.
The translation M ↦ M + 1 is a bijection between these matrices, it fixes the discriminant
t² - 4 n, and its effect on the weights cancels in total.
Very negative trace #
Reduced forms #
The elliptic fibres #
The elliptic trace fibres of Popa–Zagier's element. If t² - 4 n = -D < 0, then the
weights of the integral matrices of trace t and determinant n sum to -12 H(D): the
coefficients of Popa and Zagier's explicit element, a twelfth of the weights, sum to
-H(4 n - t²) over the matrices of trace t and determinant n.
The hypothesis D ≠ 0 is needed: for t = 2 a and n = a² with a > 0 the weights sum to
2, the weight of the scalar matrix a • 1, while -12 H(0) = 1.