3
$\begingroup$

Let D in Z/3[[x]] be sum ((a_n)(x^n)) where the sum runs over all n prime to 6 and a_n is the mod 3 reduction of the number of ideals of norm n in the ring of integers of Q(root(-3)). (So D=x+2(x^7)+2(x^13)+2(x^19)+x^25+(higher degree terms).)

There are formal Hecke operators T_p: Z/3[[x]]-->Z/3[[x]] for all primes p other than 3. Clearly the T_p with p=2 mod 3 annihilate D. Experimentally I find that they annihilate D^7, D^61 and D^547 as well.

Question: If k=(1+3^m)/4 with m odd, do the T_p with p=2 mod 3 annihilate D^k?

Remark: D is the reduction of the expansion of the modular form (eta(2z))^12, but it's not clear that this helps.

EDIT: If I've made no mistake, Noam's formula can be tweaked to show that D^k=V-U where U and V are the mod 3 reductions of the theta series attached to the binary quadratic forms 36xx+6xy+kyy and 36xx+30yy+(k+6)yy. This raises the question of which linear combos of theta series of primitive binary quadratic forms lie in the space spanned by the powers of D. I think similar questions have been answered by Nicolas and Serre (ell=2, level=1) and perhaps by Bellaiche and or Serre (ell=3, level=1).

$\endgroup$

1 Answer 1

3
$\begingroup$

Yes, and $D \equiv \eta(2z)^{12}$ certainly helps: it means $$ D^{(1+3^m)/4} \equiv \eta(2z)^{3+3^{m+1}} \equiv \eta(6z) \, \eta(2\cdot 3^{m+1}z). $$ Now $\eta = \sum_a \chi(a) q^{a^2/24}$ where $a$ runs over integers congruent to $1 \bmod 6$ and $\chi$ is the even Dirichlet character $\bmod\,12$. Since $m$ is odd, say $m=2n+1$, we have $$ \eta(6z) \, \eta(2\cdot 3^{m+1}z) = \mathop{\sum\sum}_{a,b \equiv 1 \bmod 6} \chi(ab) q^{(a^2 + 3^{2n+1}b^2)/4}. $$ All the exponents are odd, so our form is in $\ker T_2$. If $p \equiv 5 \bmod 6$ then $(a^2 + 3^{2n+1}b^2)/4$ is a multiple of $p$ iff $a,b \equiv 0 \bmod p$, whence $T_p$ multiplies our form by $p^{w-1} + 1$ where $w$ is the weight. Since $w = (1+3^m)/4$ is odd and $p \equiv 2 \bmod 3$, we have $p^{w-1}+1 = 0$ in ${\bf Z}/3{\bf Z}$ and we're done.

$\endgroup$
2
  • $\begingroup$ Thanks, Noam. The argument also occurred to me just after I submitted the question. I wonder if you have any ideas about my deeper question--is the space spanned by the D^j with j prime to 6 stable under the T_n with n prime to 6? It certainly seems true experimentally. $\endgroup$ Oct 22, 2013 at 6:52
  • $\begingroup$ Just to make the comment of mine that I'm replacing clearer: When p is 2 mod 3 and w is odd, p^(w-1)+1 is 2, not 0 in Z/3Z. So the end of Noam's proof is wrong. A revision is: ... ,whence c_(np)=c(n/p) where c_n is the coefficient of x^n in D^k. Now the coefficient of x^n in T_p(D^k) is c_(np)+p*c(n/p); since p is 2 mod 3, this is 0. (One thing to note is that in characteristics ell=2 or 3, when p isn't ell then for even w, p^(w-1)=p. So one can define formal Hecke operators on Z/ell[[x]] without taking the weight into effect, that will act well on reductions of forms of any even degree. $\endgroup$ Oct 25, 2013 at 16:44

Your Answer

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

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