Arithmetic meaning of certain periods of modular symbols of elliptic curves

92 Views Asked by At

I'm following the notation in Chapter II of Cremona's book on modular symbol algorithms.

Let $f$ be a weight two cusp form for a congruence subgroup $G$. For any two points $\alpha, \beta \in \mathcal{H}$ in the extended upper half plane equivalent under the action of $G$, there is the notion of a modular symbol $[\alpha, \beta]$, given by projecting any smooth path from $\alpha$ to $\beta$ to the quotient $X_G = \mathcal{H} / G$, which determines a closed path and hence an integral homology class in $H_1(X_G, \mathbb{Z})$. Integrating the differential $2\pi if(z)\mathrm{d}z$ over such a modular symbol yields a period of $f$, denoted by $$\langle[\alpha, \beta], f\rangle := \int_\alpha^\beta 2\pi if(z)\mathrm{d}z.$$

Now let $G$ be the congruence subgroup $\Gamma_0(N)$. For each prime $p \nmid N$, the Hecke operator $T_p$ with eigenvalue $a_p$ acts on these modular symbols in such a way that integration with $2\pi if(z)\mathrm{d}z$ yields the following identity (2.8.9): $$(1 + p - a_p) \cdot \langle[0, \infty], f\rangle = \sum_{k = 1}^{p - 1} \langle[0, \tfrac{k}{p}], f\rangle.$$

I am interested in the setting where $f$ is the weight two cusp form associated to an optimal elliptic curve $E$ over $\mathbb{Q}$ of conductor $N$. In this setting, the left hand side encodes arithmetic information of $E$: the integer $1 + p - a_p$ is just the number of points of $E$ over $\mathbb{F}_p$, while $\langle[0, \infty], f\rangle$ is the central $L$-value $L(E, 1)$ up to a sign. By dividing both sides by the real period $\Omega_E$, the factor $\langle[0, \infty], f\rangle / \Omega_E$ becomes rational, so the left hand side is also rational. In fact, the entire left hand side seemingly becomes integral, which I believe is a consequence of the optimality of $E$, possibly with further assumptions.

My question is: what is the arithmetic meaning of the individual terms $\langle[0, \tfrac{k}{p}], f\rangle / \Omega_E$ for each of $k = 1, \dots, p - 1$? The sum on the right hand side is somewhat symmetric about $k = \tfrac{p - 1}{2}$, in the sense that $\overline{\langle[0, \tfrac{k}{p}], f\rangle} = \langle[0, \tfrac{p - k}{p}], f\rangle$, so perhaps I should instead ask the question for $\Re(\langle[0, \tfrac{k}{p}], f\rangle) / \Omega_E$ for each of $k = 1, \dots, \tfrac{p - 1}{2}$. These also seem to be half-integers, which should also be a consequence of the optimality of $E$, again possibly with further assumptions. Is it possible to express these completely in terms of arithmetic invariants of $E$ without doing the standard algorithmic computation?

As an example, let $f$ be the weight two cusp form associated to the elliptic curve with Cremona label 11a1. Then $$(1 + p - a_p) \cdot \left(-\dfrac{L(E, 1)}{\Omega_E}\right) = (1 + 7 - (-2)) \cdot \left(-\dfrac{0.25384\dots}{1.2692\dots}\right) \approx -2,$$ while a period computation in SageMath shows $$\sum_{k = 1}^6 \dfrac{\langle[0, \tfrac{k}{7}], f\rangle}{\Omega_E} = 2\left(\sum_{k = 1}^3 \dfrac{\Re(\langle[0, \tfrac{k}{7}], f\rangle)}{\Omega_E}\right) = 2(\tfrac{1}{2} + \tfrac{1}{2} + (-2)) = -2,$$ as expected.

Context: weighting the periods in the right hand side appropriately gives the central $L$-value when twisted by a Dirichlet character of conductor $p$, which I am trying to understand in terms of the non-twisted central $L$-value and other arithmetic invariants. I am brand new to this area and I might be going down a bottomless rabbit hole, so any suggestions is appreciated.