Singular moduli, modular polynomials, and the index of the closure of \({\mathbb{Z}}[j(\tau)]\) in \({\mathbb{Q}}(j(\tau))\) (Q1101511)

Gross and Zagier used the prime factorization of differences of singular moduli to study modular polynomials and used the results to compute the index, I, of \({\mathbb{Z}}[j(\tau)]\) in its integral closure in \({\mathbb{Q}}(j(\tau))\). The algebraic proof they presented held only for prime discriminants. In this paper we generalize their results to relatively prime composite discriminants and give wholly algebraic proofs of our results. Let \(\tau _ 1\) and \(\tau _ 2\) be elements in the upper half plane and imaginary quadratic over \({\mathbb{Q}}\) and having fundamental discriminants \(d_ 1\) and \(d_ 2\) respectively. Let \({\mathfrak O}={\mathbb{Z}}[(d_ 1+\sqrt{d_ 1})/2]\subset {\mathbb{Q}}(\sqrt{d_ 1})=K\). Let j(z) be the elliptic modular function. The m-th modular polynomial, \(\phi _ m(x,y)\in {\mathbb{Z}}[x,y]\) is defined by \[ \phi _ m(j(z),j(z'))=\prod _{\det \gamma =m mod SL_ 2({\mathbb{Z}})}(j(z)-j(\gamma z')). \] This product is taken over all equivalence classes of \(2\times 2\) matrices of determinant m, modulo the left action of \(SL_ 2({\mathbb{Z}})\). - One simply stated result concerning \(\phi _ m\) is that any prime \(\lambda\) of either \({\mathbb{Q}}(\sqrt{d_ 1})\) or \({\mathbb{Q}}(\sqrt{d_ 2})\) dividing \(\phi _ m(j(\tau _ 1),j(\tau _ 2))\) must have characteristic \(\ell <md_ 1d_ 2/4.\) Now fix one fundamental negative discriminant d. For non-negative integers n let R(n) (resp. \(r_ 1(n))\) be the number of integral ideals (resp. integral ideals in the principal class) of \({\mathfrak O}\) having norm n. Extend both functions to \({\mathbb{R}}\) by setting \(R(x)=r_ 1(x)=0\) for arguments other than non-negative integers. For each prime p, finite of infinite, let \(\epsilon _ p: {\mathbb{Q}}_ p(\sqrt{d})\to \{\pm 1\}\) be the local character given by class field theory. - Define \(\rho _{\ell}(n)=0\) if there exist two primes \(p| d\) such that \(\epsilon _ p((n-| d|)/n)=-1\), and \(\rho _{\ell}(n)=2^{a(n)}\) otherwise, where \(a(n)=Card\{p| (n,d)\}.\) Theorem. Let \(\ell\) be a rational prime not dividing d. Then \(ord_{\ell}(I)=\sum _{n\geq 0}\sum _{k\geq 1}\rho _{\ell}(n)\cdot (R(n)-r_ 1(n))\cdot R((/d| -n)/\ell\) k).\ A more precise form of this theorem extending it to all rational primes is found in the paper.