3
$\begingroup$

Suppose that you have a generating function $$ f(q) = \sum_{k=0}^\infty a_k q^k $$ It's not too hard to obtain the generating function $$ f_{n,m}(q) = \sum_{k=0}^\infty a_{nk + m}q^k $$ by taking a creative sum of terms like $f(\zeta_n^i q)$ for appropriate roots of unity $\zeta_n$.

What about functions of the form $$ g(q) = \sum_{k=0}^\infty a_{k^2}q^k $$ or more generally, $$ g_{n,m}(q) = \sum_{k=0}^\infty a_{nk^2+m}q^k $$

Is there some way to nicely obtain these in terms of either the original generating function, or its coefficients, or anything? If it helps, the generating function that we are interested in is the inverse of the modular discriminant. That is, we are looking at $$ f(q) = \frac{1}{\Delta(q)} = q^{-1} + 24 + 324q + 3200q^2 + \cdots $$ where $\Delta(q) = q\prod_{k=1}^\infty (1-q^k)^{24} = \eta(q)^{24}$.

$\endgroup$
4
  • 1
    $\begingroup$ I don't see any reason the result should be nice in general. Consider, for example, the case $a_k = \frac{1}{k!}$. $\endgroup$ Commented Jun 12, 2014 at 23:17
  • $\begingroup$ Sometimes you can pull this out of a theta lift, e.g., if you happen to have a weight 1/2 form. See mathoverflow.net/questions/159189/… for an example. $\endgroup$
    – S. Carnahan
    Commented Jun 13, 2014 at 0:21
  • 1
    $\begingroup$ Related to the previous comment, if your generating function is a half-integer weight modular form, then the index $t$ Shimura lift will output an integer-weight modular form whose Fourier coefficients involve the numbers $a_{tk^{2}}$. It's hard to know if this is applicable without knowing where your $g(q)$ comes from. $\endgroup$ Commented Jun 13, 2014 at 12:52
  • 2
    $\begingroup$ Considering @QiaochuYuan's comment, I perhaps should have been more specific about this question, since it seems that there is no way that this will hold in full generality. $\endgroup$
    – Simon Rose
    Commented Jun 14, 2014 at 18:20

1 Answer 1

4
$\begingroup$

There is an algebraic object naturally attached to $1/\Delta$, namely the fake monster Lie algebra. It was introduced in Borcherds's paper The monster Lie algebra, but the word "fake" was later attached when Borcherds discovered a different Lie algebra that really has a well-behaved action of the monster simple group. If we forget the Lie algebra structure, we get a vector space graded by the 26-dimensional even unimodular Lorentzian lattice $I\!I_{25,1}$. The subspace whose degree is a lattice vector $v$ has dimension equal to $p_{24}(|v|^2/2+1)$, i.e., the $q^{|v|^2/2}$ coefficient of $1/\Delta$. Generating functions of the form $\sum a_{nk^2} q^k$ are then given by the Lie subalgebras supported on rays from the origin.

The phenomenon with generating functions is a special case of the Borcherds-Harvey-Moore theta lift, where a weakly holomorphic modular form of weight $1-n/2$ is sent to a meromophic automorphic form on $O(2,n)$. For $1/\Delta$, we have $n=26$, so we get an automorphic form on a rather large group. The Fourier expansion at a cusp is the somewhat unwieldy character for the fake monster Lie algebra, but we can take a one dimensional slice to get the generating function manipulation you want.

$\endgroup$
3
  • $\begingroup$ That looks really tempting. There seems to be an off-by one mismatch, (I'm really interested in the coefficients $a_{nk^2 + 1}$), but maybe a bit more care in writing my coefficients down may clear that up. $\endgroup$
    – Simon Rose
    Commented Jun 15, 2014 at 22:18
  • $\begingroup$ You could in theory use a ray that is not based at the origin, but you no longer get a Lie subalgebra. $\endgroup$
    – S. Carnahan
    Commented Jun 15, 2014 at 22:20
  • $\begingroup$ I'd thought of that, but I'm hoping that it's just an off-by-one error, since I'm guessing that there are nicer properties if it arises due to a subalgebra. $\endgroup$
    – Simon Rose
    Commented Jun 15, 2014 at 22:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .