6
$\begingroup$

Let $f(z) = \sum_{n=0}^\infty a_nq^n$ be the fourier expansion of a (quasi-)modular form (with $q = e^{2\pi i z}$). Consider the following related functions:

$$f_{m,k}(z) = \sum_{n=0}^\infty a_{mn + k}q^n$$

$$g_{m,k}(z) = \sum_{n=0}^\infty a_{mn + k}q^{mn+k}$$

for, naturally, $0 \leq k < m$.

My question: are these in any way modular? For some congruence subgroup, etc. etc.?

What if we put some restrictions on $f(z)$? Or shift the powers of $q$ somewhat?

As an example of when we do retain (quasi-)modularity, consider the case $f(z) = E_2(z), m = 2$. Then it is not too hard to show that

$$ E_2(z) = g_{2,1}(z) + 3E_2(2z) - 2E_2(4z) $$

and so it follows that $g_{2,1}(z)$ is indeed quasi-modular for $\Gamma_0(4)$. Since then $g_{2,0} = E_2(z) - g_{2,1}(z)$, this is also quasi-modular.

So is this in general true?

$\endgroup$

1 Answer 1

3
$\begingroup$

I know nothing about quasimodular forms, so let me answer the question for modular forms in the classical sense.

First, your $f_{m, k}$. If $k = 0$ this is essentially the image of $f$ under the Hecke operator $U_m$, so it is always going to be modular of some level (and if f has level $\Gamma$, then $f_{m, 0}$ will have level $ \Gamma \cap \Gamma_0(m)$ or something like that). If $k \ne 0$ then $f_{m, k}$ will not be modular, because $f_{m, k}(mz) = q^{-k} g_{m, k}(z)$, $g_{m, k}$ is modular, and dividing by a power of $q$ destroys modularity.

Now $g_{m, k}$. This is just a linear combination of the forms $f(z + j/m)$ for $j = 0, ..., m$ by finite Fourier theory, and all these are modular, so $g_{m, k}$ is modular (although if $g$ is modular of level $\Gamma_1(N)$, then $g_{m, k}$ will have level something like $\Gamma_1(mN) \cap \Gamma_0(m^2 N)$ in general).

The forms $g_{m, k}$ are closely related to the forms $$ g_{\chi} = \sum_n \chi(n) a_n q^n $$ for Dirichlet characters $\chi$ modulo $m$, which are very interesting (and well-studied) objects; if $g$ is a Hecke eigenform then so are the forms $g_{\chi}$ (although the $g_{m, k}$ generally aren't).

$\endgroup$
4
  • $\begingroup$ But can't one write the $g_{m,k}$ as linear combinations of forms of the form $g_\chi$ (for various Dirichlet characters $\chi$ modulo divisors of $m$)? $\endgroup$
    – ACL
    Nov 21, 2013 at 17:48
  • $\begingroup$ The expression for $g_{m,k}$ as a linear combination of the $g_\chi$ is given in Soma Purkait's thesis (page 68). She also determines the level. wrap.warwick.ac.uk/50236 $\endgroup$
    – Siksek
    Nov 21, 2013 at 20:03
  • $\begingroup$ Ah, this is good news; as it turns out I don't think I actually need the functions $f_{m,k}(q)$ that I mentioned. $\endgroup$
    – Simon Rose
    Nov 21, 2013 at 21:32
  • $\begingroup$ @SamirSiksek: That seems to be almost exactly what I want, although that setting seems to be more restrictive (e.g. gcd(k,m) = 1, and the original function must be a cusp form). Can this restriction be removed? Or does that proof simply apply in that restricted case? $\endgroup$
    – Simon Rose
    Nov 21, 2013 at 21:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.