5
$\begingroup$

The group $\operatorname{SL}_2(\mathbb Z)$ acts on polynomials in two variables $\mathbb C[x,y]$ via $A\cdot f(x,y)\mapsto f(A^{-1}.(x,y))$ where $(x,y)$ is regarded as a column vector. There are two standard matrices $S=\begin{bmatrix}0&-1\\1&0\end{bmatrix}$ and $T=\begin{bmatrix}1&1\\0&1\end{bmatrix}$ which generate $\operatorname{PSL}_2(\mathbb Z)$. Let $\epsilon = \begin{bmatrix}-1&0\\0&1\end{bmatrix}\in \operatorname{GL}_2(\mathbb Z)$.

In the course of studying the Johnson homomorphism, I defined a space $\Omega_2(V)$ whose specialization to the case $V=\mathbb C$ in degree $2n$ is defined as follows. Let $R=\mathbb C[x,y]_{2n}$ and let $\langle a_1,\ldots, a_m\rangle=a_1R+\cdots+a_mR$. $$\Omega_2(\mathbb C)_{2n}=\dfrac{R}{\langle 1+ST+(ST)^2,1-\epsilon S\rangle+\mathbb C\{x^{2n},y^{2n}\}}.$$

Computer calculations give the dimensions as $0,1,1,2,3,3,4,5,5,6,\ldots$ starting at $n=1$, and I am looking for a proof of the claim that this pattern continues.

If you add the additional relation $\langle 1-\epsilon\rangle$ to this presentation, then it is not hard to see, using for example modular symbols, that the quotient is isomorphic to $H^1(\operatorname{GL}_2(\mathbb Z);R)$, which is known to be isomorphic to the space of cusp forms for $\operatorname{SL}_2(\mathbb Z)$ of weight $2n+2$. So we have an exact sequence: \begin{multline*}0\to\dfrac{\langle 1-\epsilon\rangle}{\langle 1+(ST)+(ST)^2,1-\epsilon S\rangle\cap \langle 1-\epsilon\rangle+\mathbb C\{x^{2n},y^{2n}\}}\\ \to \Omega_2(\mathbb C)_{2n}\to H^1(\operatorname{GL}_2(\mathbb Z);R)\to 0\end{multline*} All the dimensions work out correctly if $$\langle 1+(ST)+(ST)^2,1-\epsilon S\rangle\cap\langle 1-\epsilon\rangle= \langle 1-\epsilon S\rangle\cap\langle 1-\epsilon\rangle=\mathbb C\{x^{2m}y^{2n-2m}-x^{2n-2m}y^{2m}\,:\, 0\leq m\leq n\},$$ but I have not been able to prove this!

Remarks:

  • It might help to use the fact that $\langle 1+(ST)+(ST)^2\rangle=\ker(1-(ST))$.

  • I asked an equivalent version of this question at math.stackexhange, but have gotten no useful responses.

  • The dimensions of $\Omega_2(\mathbb C)$ resemble those of the space of cusp forms for the congruence subgroup $\Gamma_0(3)$, but the polynomial degrees are off by one, and so it seems unlikely that they are related.

$\endgroup$

1 Answer 1

3
$\begingroup$

Martin Kassabov sent me an elegant solution to this problem. Start with the observation that $\epsilon S$ and $ST$ generate a copy of the symmetric group $S_3$. Now decompose $R$ into a direct sum of irreducible $S_3$-modules. Modding out by $1+ST+(ST)^2$ and $1-\epsilon S$ kill both $1$ dimensional representations and reduce the dimension of the $2$ dimensional representation to $1$. So $\Omega_{2n}(\mathbb C)$, if you ignore the $x^{2n},y^{2n}$, will have the same dimension as the multiplicity of the two dimensional representations in $R$. It's also not hard to show that further modding out by these two monomials will reduce the dimension by $1$. Now to calculate the $S_3$ decomposition, calculate the character. It's not too hard to see that $\chi_R=\begin{bmatrix} 2n+1&1&1\end{bmatrix}, \begin{bmatrix} 2n+1&1&-1\end{bmatrix}$ or $\begin{bmatrix} 2n+1&1&0\end{bmatrix}$ depending on the congruence class of $n$ modulo $3$. From here it follows that the multiplicity of the 2D representation in $R$ is $\frac{2n+1}{3}, \frac{2n+2}{3}$ or $\frac{2n}{3}$ again depending on the congruence class modulo 3.

$\endgroup$

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.