0
$\begingroup$

I'm missing something obvious, I suspect, but for the life of me I can't figure it out. What is wrong with the following exposition?


For any integral matrix $α$ with positive determinant let $[α]_k$ be the linear operator on the space of holomorphic functions on the extended upper-half plane $H^* = H ∪ \{i∞\}$ defined by $$f[α]_k(τ) = (\det α)^{k/2} j_α(τ) f(ατ)$$ where $j_α$ is the typical $k^{\text{th}}$ factor of automorphy of $α$ and $α$ acts on $H^*$ by usual fractional linear transformation. Then so long as $f$ is a modular form (holomorphic on $H^*$ plus $f[α]_k = f$ for all $α$ with determinant $1$) then we see $f[α]_k = f$ for all $α$ with positive determinant.

Now, consider the Hecke operator $T(n)$ of order $n$, which acts on the free abelian group of lattices via $$T(n)L = \sum_{[L:L'] = n} L'$$ Using our knowledge of bases of free abelian groups of finite rank, this equates to $$\sum αL$$ where $α$ runs over a set of representatives of the set of matrices of determinant $n$ under the action of left multiplication by $\text{SL}_2(ℤ)$ (which we know there to be $σ(n)$ of, when $σ(n)$ is the sum of the divisors of $n$). But if $F$ is the corresponding function on lattices obtained from $f$, then $$F(α[1,τ]) = F([a+bτ, c+dτ]) = F((a+bτ)[1,\dfrac{c+dτ}{a+bτ}]) =$$$$(a+bτ)^{-k}F([1,\dfrac{c+dτ}{a+bτ}]) = n^{-k/2} f[\rho α \rho^{-1}]_k(τ)$$ where $\rho = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}$.

As such, if we convert the action of $T(n)$ to one operating on the space of modular forms of weight $k$ I should expect to calculate (where $τ∈H^*$ and $L = [1,τ]$) $$T(n)f(τ) = F(T(n)L) = F(\sum αL) = \sum F(αL) =$$$$\sum n^{-k/2} f[\rho α \rho^{-1}]_k(τ) = \sum n^{-k/2} f(τ) = n^{-k/2} σ(n) f(τ)$$

Yet this is obviously false, because $f$ was arbitrary yet I know not every modular form is a Hecke eigenform, and also that the eigenvalue $n^{-k/2} σ(n)$ is incorrect even when $f$ is an eigenform.

1 Answers 1

1

Nevermind; despite several hours pulling my hair out before asking the question, I finally found the fallacy; $f[α]_k = f$ for all matrices of determinant $1$ would've implied $f[α]_k = f$ for all integral matrices of positive determinant, but dividing an integral matrix by the square root of its determinant may not be an integral matrix, as was implicit in my argument.

Unsure whether to delete or close this thread, if necessary.