Variants of Lehmer's speculation for newforms (Q6110338)

Ramanujan's tau-function, \(\tau(n)\), is defined by \(\Delta(z)=\sum_{n=1}^{\infty} \tau(n)q^{n}=q \prod_{n=1}^{\infty} \left( 1- q^{n} \right)^{24}\). It has been the subject of much study since first introduced by Ramanujan. Much of its importance arises from the fact that \(\Delta(z)\) is a modular form (a cusp form of weight \(12\) and level \(1\), in fact). \textit{D. H. Lehmer} [Duke Math. J. 14(2), 429--433 (1947; Zbl 0029.34502)] famously asked if \(\tau(n)\) can ever be zero for \(n>0\). This question remains open, although it is known that any counterexamples must satisfy \(n>8.16 \cdot 10^{23}\) (see the unpublished paper [\textit{M. Derickx} et al., ``Computing Galois representations and equations for modular curves \(X_H(\ell)\)'', Preprint, \url{arXiv:1312.6819}]). One can ask about other possible values of \(\tau(n)\) too. Furthermore, one can also consider this question for the Fourier coefficients of other modular forms. This is the subject of this paper, which has already led to further significant further work by other mathematicians (see the remark at the bottom of page~3 and top of page~4 for details and references). The authors provide a method for effectively answering this question for odd values of the Fourier coefficients (see their Theorem~1.1) for a large class of modular forms. Note that Theorem~1.1 and its generalisation, Theorem~3.2, are stated for values that are odd prime powers in absolute value, but such results can be readily extended to any odd integer values by Hecke multiplicativity. Applying this method to \(\tau(n)\), they show that for \(n>1\), \(\tau(n) \not\in \left\{ \pm 1, \pm 3, \pm 5, \pm 7, \pm 13, \pm 17, -19, \pm 23, \pm 37, \pm 691 \right\}\), and even more values under GRH (see their Theorem~1.2). The use of GRH here is only needed to make the solution of the Thue equations that arise feasible. As mentioned above, their method applies more generally. It can be used for any even weight (at least \(4\)) newform with integer coefficients and trivial mod \(2\) residual Galois representation. Their Theorem~3.2 is the generalisation of Theorem~1.1 to such newforms and their Theorem~1.3 provides a generalisation of Theorem~1.2. They also consider some related questions. One such question is whether a fixed odd value can be a Fourier coefficient of newforms with arbitrarily large weight. Their Theorem~1.4 states that this is not the case. Moreover, their bound on the size of the weight is effective. Another is about lower bounds for the number of prime divisors of coefficients of newforms. This is addressed in their Theorem~1.5 for the Ramanujan \(\tau\)-function and Theorem~2.5, more generally. The bound in Theorem~1.5 is best possible, as prime values of \(\tau(n)\) show (for example, Lehmer's example \(\tau \left( 251^{2} \right)=-80561663527802406257321747\)). The main tool in their proofs is the result of \textit{Yu. Bilu} et al. [J. Reine Angew. Math. 539, 75--122 (2001; Zbl 0995.11010)] on primitive prime divisors of Lucas sequences. The connection with the problems investigated here is that the theory for Lucas sequences applies to the recursion relations given by Hecke operators in the theory of modular forms. The Chabauty-Coleman method and results on Thue equations also play roles in the proof.