Refined dimensions of cusp forms, and equidistribution and bias of signs (Q1708064)

The paper under review gives dimension formulas for various subspaces of cusp forms of squarefree level, as well as interesting consequences. Let \(N>1\) be squarefree and \(S^{\mathrm{new}}_k(N)\) be the usual space of newforms on \(\Gamma_0(N)\) of weight \(k\). It is well known that \(S^{\mathrm{new}}_k(N)\) is a direct sum of subspaces \(S^{\mathrm{new},\pm}_k(N)\), which consist of newforms with global root number \(\pm1\). The author proves dimension formulas for these plus and minus spaces (see Theorem 2.2), based on the relation \(\dim S^{\mathrm{new},\pm}_k(N)=\frac{1}{2}(\dim S^{\mathrm{new}}_k(N)\pm \mathrm{tr}_{S^{\mathrm{new}}_k(N)}W_N)\), the trace formula of \textit{M. Yamauchi} [J. Math. Kyoto Univ. 13, 403--411 (1973; Zbl 0267.10038)], and \textit{G. Martin}'s formula for \(\dim S^{\mathrm{new}}_k(N)\) [J. Number Theory 112, No. 2, 298--331 (2005; Zbl 1095.11026)]. Here \(W_N\) is the product of the \(p\)-th Atkin-Lehner operators \(W_p\) with \(p\mid N\). Interesting consequences include that \(\dim S^{\mathrm{new},+}_k(N)\geq\dim S^{\mathrm{new},-}_k(N)\); more precisely, Corollary 2.3 shows that for bounded weights and fixed level the global root numbers are biased toward \(+1\), while the global root numbers are equidistributed for large weight in the following sense: \[ \dim S^{\mathrm{new},\pm}_k(N)=\frac{(k-1)\varphi(N)}{24}+O(2^{\omega(N)}),\quad\text{as }k\rightarrow\infty. \] The error term in the above greatly improves on the asymptotic (2.73) in [\textit{H. Iwaniec} et al., Publ. Math., Inst. Hautes Étud. Sci. 91, 55--131 (2000; Zbl 1012.11041)]. On the other hand, \(S^{\mathrm{new}}_k(N)\) has finer subspaces \(S^{\mathrm{new},\varepsilon_M}_k(N)\) with a certain sign pattern \(\varepsilon_M\) for \(M\mid N\) (see the second paragraph of Section 3 for definition). The author establishes an explicit formula for \(\dim S^{\mathrm{new},\varepsilon_M}_k(N)\) in Theorem 3.3. Again there are several interesting corollaries, among which we mention Corollary 3.6. This particular result shows perfect unbiasedness of sign patterns if the squarefree level \(N\) is ``large'', roughly in the sense that \(N\) can also be divisible by a certain set of odd primes coprime with \(M\) (see Corollary 3.6 for details). In addition, the author obtains results on Galois orbits in \(S^{\mathrm{new}}_k(N)\). We mention Proposition 4.1, which basically says, with \(1<M\mid N\) squarefree, there is an explicit number \(K_{N,M}\) such that for any even weight \(k>K_{N,M}\) one has that all possible sign patterns for \(M\) will occur for \(S^{\mathrm{new}}_k(N)\), and there are at least \(2^{\omega(M)}\) Galois orbits in \(S^{\mathrm{new}}_k(N)\).