Loading [MathJax]/extensions/TeX/boldsymbol.js
Skip to main content

Section 5.2 Modular forms (Asra)

Last time we saw the j-function, which was \SL_2(\ZZ)-invariant, this is quite a strong condition, and in fact j is pretty much all we get under this condition. So instead we weaken this somewhat to some other variance property.

If w = f(z) \diff z on \HH and f(z) is meromorphic. \gamma \in \Gamma then

\begin{equation*} f(\gamma z) d(\gamma z) = f(\gamma z) d\left( \frac{az+b}{cz+d} \right) \end{equation*}
\begin{equation*} = f(\gamma z) \left( \frac{\cdots \diff }{(cz+d)^2} \right) \end{equation*}

so we get a condition

\begin{equation*} f(\gamma z) = (cz+d)^2 f(z) \end{equation*}

this is how we come to:

Definition 5.2.1.

A holomorphic function f\colon \HH \to \CC is a weakly modular function for \Gamma of weight k if

\begin{equation*} f(\gamma z) = (cz+d)^k f(z) \forall \gamma = \begin{pmatrix} a\amp b \\ c \amp d \end{pmatrix} \in \Gamma\text{.} \end{equation*}
Remark 5.2.2.

If -I \in \Gamma and k odd

\begin{equation*} f(-z) = - f(-z) \end{equation*}

so in this setting we only have interesting behaviour for even k\text{.}

If \Gamma is a congruence subgroup of level N we have

\begin{equation*} \begin{pmatrix} 1\amp N \\ 0 \amp 1 \end{pmatrix} \in \Gamma \end{equation*}

gives you a q-expansion

\begin{equation*} q= e^{2\pi i z} \end{equation*}
\begin{equation*} f(z) = \sum_{m\in \ZZ} a_m q^{m/N}\text{.} \end{equation*}

f is holomorphic at \infty if a_m = 0 for m \lt 0\text{.}

f is holomorphic at all cusps if f(\gamma z)(cz+d)^k is holomorphic at \infty for all \gamma \in \SL_2(\ZZ)\text{.}

Example 5.2.3.

Cusps for \Gamma_0(p)\text{,} we know we have \infty\text{,} what is the orbit of this?

\begin{equation*} \gamma\in \Gamma_0(p),\, \gamma = \begin{pmatrix} a\amp b \\ cp \amp d \end{pmatrix} \end{equation*}
\begin{equation*} \gamma \infty = \frac{a}{cp} \end{equation*}

so anything with a p in the denominator is equivalent to \infty\text{,} what about the rest?

\begin{equation*} \gamma 0 = \frac{b}{d}, \, \gcd(b,d) =1\text{,} \end{equation*}

so we have two cusps.

Definition 5.2.4. Modular forms.

A modular form is a weakly modular function that is holomorphic at all the cusps.

Example 5.2.5.

Eisenstein series

\begin{equation*} G_k(z) = \sum'_{m,n\in \ZZ} \frac{1}{(mz+n)^k} \end{equation*}

is a modular form of weight k \gt 2 for \SL_2(\ZZ)\text{.}

\begin{equation*} \lim_{\im z\to \infty} G_k(z) = \lim_{\im z\to \infty} \sum'_{m,n\in \ZZ} \frac{1}{(mz+n)^k} = \sum'_{n\in \ZZ} \frac{1}{n^k} = 2\zeta(k)\text{.} \end{equation*}

So here the function does not vanish at 0.

Definition 5.2.6. Cusp forms.

A cusp form is a modular form that vanishes at all cusps.

Given a cusp it will be stabilised by some

\begin{equation*} \begin{pmatrix} 1\amp h \\ 0 \amp 1 \end{pmatrix} \end{equation*}

call the smallest such h for a given cusp the width of the cusp.

Example 5.2.7.

Let's find the width of a cusp in \Gamma_0(qp) we have

\begin{equation*} \begin{pmatrix} 1\amp 1 \\ 0 \amp 1 \end{pmatrix} \end{equation*}

so the width of \infty is 1.

What about \alpha = 1/p\text{?}

  1. Find an element \gamma \in \SL_2(\ZZ) s.t. \gamma(\infty) = \alpha\text{.}
  2. Compute
    \begin{equation*} \delta(x) = \gamma \begin{pmatrix} 1\amp x \\ 0 \amp 1 \end{pmatrix} \gamma\inv \end{equation*}
  3. Find the smallest x such that \delta(x) = \Gamma_0(pq)
\begin{equation*} \gamma = \begin{pmatrix} 1\amp 0 \\ p \amp 1 \end{pmatrix},\, \gamma(\infty) = \frac 1p \end{equation*}
\begin{equation*} \begin{pmatrix} 1\amp 0 \\ p \amp 1 \end{pmatrix} \begin{pmatrix} 1\amp x \\ 0 \amp 1 \end{pmatrix} \begin{pmatrix} 1\amp 0 \\ -p \amp 1 \end{pmatrix} = \begin{pmatrix} 1 - px \amp x \\ -p^2 \amp px+ 1 \end{pmatrix} \end{equation*}
Example 5.2.8. A cusp form.

Let \Delta(\tau) = g_2(\tau)^3 - 27g_3(\tau)^2,\,g_2(\tau) = 60G_4(\tau),\,g_3(\tau) =140G_6(\tau) \Delta(\tau) has weight 12 for \SL_2(\ZZ)\text{.} This vanishes at \infty because

\begin{equation*} \zeta(4) = \frac{\pi^4}{90} \end{equation*}
\begin{equation*} \zeta(6) = \frac{\pi^6}{945} \end{equation*}

also

\begin{equation*} j(z) = \frac{g_2(\tau)^3}{\Delta(\tau)} \end{equation*}

so \Delta(\tau) vanishes at \infty because g_2(\tau) doesn't and j(z) has a simple pole at \infty\text{.}

M_k(\Gamma) as the space of modular forms of weight k for \Gamma is a \CC -v.s. S_k(\Gamma) as the space of cusp forms of weight k for \Gamma is a \CC -v.s.

\begin{equation*} \dim(S_2(\Gamma)) = g(X(\Gamma)) \end{equation*}

Given an elliptic curve

\begin{equation*} E/ \CC = \CC/\Lambda \end{equation*}
\begin{equation*} E \to E' \end{equation*}
\begin{equation*} \CC/\Lambda \to \CC/\Lambda'\text{,} \end{equation*}

studying degree n isogenies, is like studying index n sublattices

Definition 5.2.11. Hecke operators.

n \ge 1 then T(n) is the nth Hecke operator acting on

\begin{equation*} \divisors (\mathcal L) \end{equation*}

by

\begin{equation*} T(n) \Lambda = \sum_{\Lambda' \subseteq \Lambda,\,[\Lambda : \Lambda'] = n} (\Lambda') \end{equation*}
Definition 5.2.12.

Let \lambda \in \CC^\times the homothety operator R_\lambda is R_\lambda \Lambda = \lambda\Lambda\text{.}

Of 4.

\(\Lambda \in \mathcal L\) for \(\Lambda' \subseteq \Lambda \) index \(p^{e+1}\) have

\begin{equation*} a(\Lambda') = \# \{ \Gamma : \Lambda' \subseteq \Lambda \subseteq_p \Lambda \} \end{equation*}
\begin{equation*} b(\Lambda') = 1 \text{ if } \Lambda' \subseteq p\Lambda \end{equation*}

now

\begin{equation*} T(p^e)T(p) \Lambda = T(p^e) \sum_{\Gamma \subseteq_p \Lambda } (\Gamma) = \sum_{\Gamma \subseteq_p \Lambda } \sum_{\Lambda' \subseteq_{p^e} \Gamma} (\Lambda') =\sum_{\Lambda' \subseteq_{p^e} \Gamma} a(\Lambda') (\Lambda') \end{equation*}
\begin{equation*} T(p^{e+1} ) \Lambda = \sum_{\Lambda'\subseteq_{p^{e+1}} \Lambda} (\Lambda') \end{equation*}
\begin{equation*} T(p^{e-1} )R_p \Lambda = T(p^{e-1}) (p \Lambda) = \sum_{\Lambda'' \subseteq_{p^{e-1}} p\Lambda } (\Lambda'') = \sum_{\Lambda' \subseteq_{p^{e+1}} \Lambda} b(\Lambda') (\Lambda') \end{equation*}

Split into cases, do some maths..

Hecke operators on lattices Given \Lambda' \subseteq_n \Lambda there is an integer matrix of determinant n taking one basis to the other. Have a correspondence

\begin{equation*} \{ \alpha \in M_2(\ZZ) : \det(\alpha) = n \} \leftrightarrow \{ \Lambda ' : \Lambda ' \subseteq_n \Lambda\} \end{equation*}

representatives in Hermite normal form

\begin{equation*} S_n = \{\begin{pmatrix} a\amp b \\ 0 \amp d\end{pmatrix} : ad= n,\, a,d\gt 0\, 0\le b \lt d\} \end{equation*}

The Hecke operators act on modular forms f(\tau) by reinterpreting weakly modular functions of weight k as functions on lattices that have a weight k action under homothety.

This boils down to

\begin{equation*} (T_k(n) f)(\tau) = n^{k-1} \sum_{ ad= n,\, a,d\gt 0\, 0\le b \lt d} d^{-k} f\left( \frac{ a\tau +b}{d}\right) \end{equation*}

We have an action on fourier expansions

\begin{equation*} f(\psi) = \sum_{m\in \ZZ} a_m q^m \end{equation*}
\begin{equation*} T_k(p)f(\tau) = p^{k-1} \sum_{m\in \ZZ} a_m q^{pm} + \frac{1}{p} \sum_{b=0}^{p-1} \left( \sum_{m\in \ZZ} a_m e^{2\pi i m (z+b)/p}\right) \end{equation*}
\begin{equation*} = p^{k-1} \sum_{m\in \ZZ} a_m q^{pm} + \frac{1}{p} \sum_{m\in \ZZ} a_me^{2\pi i m z/p} \sum_{b=0}^{p-1} \underbrace{e^{2\pi i m b/p}}_{p\text{ if }p|m,0\text{ otw}} \end{equation*}
\begin{equation*} = p^{k-1} \sum_{m\in \ZZ} a_m q^{pm} +\sum_{m\in \ZZ} a_{pm}q^m \end{equation*}

If f\in S_k(\Gamma_0(1)) is an eigenfunction for these operators we can normalise so that a_1(f) = 1\text{.}

\begin{equation*} T(m)T(n) = T(mn) \end{equation*}
\begin{equation*} a_ma_n = a_{mn} \end{equation*}
\begin{equation*} a_{p^r} = a_p a_{p^{r-1}} + p^{k-1} a_{p^{r+1}} \end{equation*}
Definition 5.2.18. Petersson inner product.

The Petersson inner product of two cusp forms f,g\in S_k(\SL_2(\ZZ)) is defined to be

\begin{equation*} \pair{f}{g} = \int_{\mathcal D} f \bar g y^{k-2} \diff x \diff y \end{equation*}

where \mathcal D is a fundamental domain for \SL_2(\ZZ)\text{.}