A remark on the values of the zeta functions associated with cusp forms (Q790156)

Let \(f(z)=\sum_{n=1}a(n)e(nz)\) and \(g(z)=\sum_{n=1}b(n)e(nz)\) be primitive cusp forms (normalized newforms) of conductor N, where \(e(z)=\exp(2\pi iz)\) and z is in the upper half plane. Let k and \(\ell\) be the weights, \(\chi\) and \({\bar \chi}\) the characters of f and g, respectively. Suppose that \(k>1\). Define a zeta-function \(D(s,f,g)=\sum_{n=1}a(n)b(n)n^{-s}\) (with a complex variable s), and let K be the field generated over \({\mathbb{Q}}\) by a(n) and b(n). \textit{G. Shimura} [Commun. Pure Appl. Math. 29, 783-804 (1976; Zbl 0348.10015)] proved that the special values \(\pi^{-k} <f,f>^{-1} D(m,f,g)\) belong to K for integers m with \((k+\ell -2)/2<m<k,\) where \(< , >\) denotes the normalized Petersson product. The author proves that (the numerator ideals of) these special values are divisible by a certain polynomial of a(p) and b(p) for prime divisors p of N. This is also true, in a modified form, if g is replaced by an Eisenstein series. The precise result is too complicated to be stated here. It is illuminated by some interesting examples for levels \(N=13\) and \(N=5\).