Zagier duality for level \(p\) weakly holomorphic modular forms (Q2337345)

Don Zagier proved the following duality result (see Theorem 4 in [\textit{D. Zagier}. Motives, polylogarithms and Hodge theory. Part I: Motives and polylogarithms. Somerville, MA: International Press. 211--244 (2002; Zbl 1048.11035)]). For any integer \(k\), let \(M^{!}_{k+1/2}\) be the space of holomorphic functions \(f\) on the upper half plane \(\mathcal{H}\) such that \(f\) transforms under the action of \(\Gamma_0(4)\) like \(\theta^{2k+1}\) where \(\theta\) is the Jacobi function \(\theta=\sum_{n\in\mathbb{Z}}q^{n^2}\), (\(q=\exp(2\pi i\tau)\)), \(f\) is meromorphic at the cusps and its Fourier expansion at \(\infty\) has the form \(\sum_nc(n)q^n\) with \(c(n)=0\) if \((-1)^kn\not\equiv0,1\pmod{4}\). This space \(M^{!}_{k+1/2}\) is the Kohnen plus-subspace of the space of weakly holomorphic modular forms of weight \(k+1/2\) over \(\Gamma_0(4)\). It has infinite dimension. For every integer \(d\geq0\) with \(d\equiv 0,3\pmod{4}\). Let \(f_d\) be the unique form in \(M^{!}_{1/2}\) having a Fourier expansion of the shape \[ f_d(\tau)=q^{-d}+\sum_{D>0}A(D,d)q^D. \] A basis of \(M^{!}_{1/2}\) is \(\{f_d\colon d\geq0,\, d\equiv 0,3\pmod{4} \}\). The coefficient \(A(1,d)\) is the value at \(d\) of the modular trace function: a weighted version for the trace of the values taken by the modular invariant \(j\) function at elements in \(\mathcal{H}\) satisfying a quadratic equation over \(\mathbb{Z}\). For each integer \(D>0\) congruent to \(0\) or \(1\) modulo \(4\), let \(g_D\) be the unique form in \(M^{!}_{3/2}\) having a Fourier expansion of the form \[ g_D(\tau)=q^{-D}+\sum_{d\geq 0}B(D,d)q^d. \] The coefficient \(B(1,d)\) is the opposite of the value at \(d\) of the modular trace function. Zagier duality theorem is the equality \(A(D,d)=-B(D,d)\) for all \(D\) and \(d\). The proof is obtained by computing the generating series of both families \(\{f_d\colon d\geq0,\, d\equiv 0,3\pmod{4} \}\) and \(\{g_D\colon D>0,\, d\equiv 0,1\pmod{4} \}\). This duality theorem has been extended by \begin{itemize} \item \textit{J. Rouse} (weakly holomorphic modular forms with characters of weight \(0\) or \(2\) and level \(5\), \(13\) or \(17\) in [J. Lond. Math. Soc., II. Ser. 73, No. 2, 339--354 (2006; Zbl 1161.11335)]), \item \textit{D. Choi} (weakly holomorphic modular forms with characters of weight \(0\) or \(2\) and prime level in [Proc. Am. Math. Soc. 134, No. 12, 3445--3447 (2006; Zbl 1163.11323)]), \item \textit{K. Bringmann} and \textit{K. Ono} (half-integral weight Maass-Poincaré series for \(\Gamma_0(4)\) in Kohnen's plus-space, in [Math. Ann. 337, No. 3, 591--612 (2007; Zbl 1154.11015)]), \item \textit{W. Duke} and \textit{P. Jenkins} (weakly holomorphic modular forms of integral weight and level \(1\) in [Pure Appl. Math. Q. 4, No. 4, 1327--1340 (2008; Zbl 1200.11027)]), \item \textit{B. Cho} and \textit{Y. Choie} (vector-valued harmonic weak Maass forms of integral weight in [Proc. Am. Math. Soc. 139, No. 3, 787--797 (2011; Zbl 1277.11039)]), \item \textit{Y. Zhang} (weakly holomorphic modular forms of integral weight, under the assumption that the \(2\)-part of the level of the forms is \(1\) or \(4\) in [J. Number Theory 155, 139--162 (2015; Zbl 1414.11056)]) \end{itemize} among others. In this article, the authors provide duality results for the subspace of weakly holomorphic modular forms of integral weight consisting of weakly holomorphic modular forms of integral weight that are holomorphic at every cusps different from \(\infty\). For any integer, the number of cusps of \(\Gamma_0(N)\) is \(\sum_{d\mid N}\varphi\left(\gcd(d,N/d)\right)\). If \(N>1\), at least one cusp is not \(\infty\). Let \(\mathcal{C}_0(N)\) be the set of cusps of \(\Gamma_0(N)\) that are not \(\infty\). Let \(M^{\#}_k(N)\) be the space of weakly holomorphic modular forms of weight \(k\) over \(\Gamma_0(N)\) that are holomorphic on \(\mathcal{C}_0(N)\), and \(S^{\#}_k(N)\) its subspace of forms that vanish on \(\mathcal{C}_0(N)\). Let \(\{f_{k,m}^{(N)} : m\in\mathcal{M}_k^{(N)}\}\) be the reduced row echelon basis for \(M^{\#}_k(N)\), whose elements have a Fourier expansion is of the shape \(q^{-m}+\sum_{n>-m}a_k^{(N)}(m,n)q^n\). The coefficient \(a_k^{(N)}(m,n)\) is set to \(0\) if there is no coefficient for \(q^n\) is the expansion of \(f_{k,m}^{(N)}\) and, \(a_k^{(N)}(m,-m)=0\) by convention. We shall they that there is a gap in the basis if the set \(\mathcal{M}_k^{(N)}\) is not a set of consecutive elements. Similarly, let \(\{g_{k,m}^{(N)} : m\in\mathcal{S}_k^{(N)}\}\) be the reduced row echelon basis for \(S^{\#}_k(N)\), whose elements have a Fourier expansion is of the shape \(q^{-m}+\sum_{n>-m}b_k^{(N)}(m,n)q^n\). The coefficient \(b_k^{(N)}(m,n)\) is set to \(0\) if there is no coefficient for \(q^n\) is the expansion of \(g_{k,m}^{(N)}\) and, \(b_k^{(N)}(m,-m)=0\) by convention. The authors proves the duality \(a_k^{(p)}(m,n)=-b_{2-k}^{(p)}(n,m)\) for all \(m\) and \(n\) in \(\mathbb{Z}\) in the following cases : \begin{itemize} \item for any prime level \(p\not\equiv 1\pmod{12}\) such that \(p\geq 11\) (or, equivalently, such that the genus of \(\Gamma_0(p)\) is positive), and any integer weight \(k\in 2\mathbb{Z}\) such that \(p-1\) divides \(k\) or \(k-2\), \item for any prime level \(p\in\{11,17,19,23,29,31,37\}\) and any weight \(k\in 2\mathbb{Z}\). \end{itemize} Finally, the authors derive from the previous results a closed form for the generating function of \(f_{k,m}^{(p)}\) for \(p\in\{11,17,19\}\) and arbitrary weight \(k\in 2\mathbb{Z}\). If there is no gap in the basis \(\{f_{k,m}^{(N)} : m\in\mathcal{M}_k^{(N)}\}\), the three following functions are equal: \[ \sum_mf_{k,m}^{(N)}(\tau)\exp(2\pi imz), \] \[ \frac{\left(a_{0}^{(p)}(2,-1)g_{2-k,n_0+1}^{(p)}(z)+g_{2-k,n_0+2}^{(p)}(z)\right)f_{k,-n_0}^{(p)}(\tau)+g_{2-k,n_0+1}^{(p)}(z)f_{k,-n_0+1}^{(p)}(\tau)}{f_{0,2}^{(p)}(z)-f_{0,2}^{(p)}(\tau)} \] and \[ \frac{\left(b_{0}^{(p)}(2,-1)f_{k,-n_0}^{(p)}(\tau)+f_{k,-n_0+1}^{(p)}(\tau)\right)g_{2-k,n_0+1}^{(p)}(z)+f_{k,-n_0}^{(p)}(\tau)g_{2-k,n_0+2}^{(p)}(z)}{g_{0,2}^{(p)}(z)-g_{0,2}^{(p)}(\tau)} \] where \(n_0\) is the largest element in \(\mathcal{M}_k^{(N)}\) (equivalently, the opposite of the valuation of the generating series). If there is a gap in the basis \(\{f_{k,m}^{(N)} : m\in\mathcal{M}_k^{(N)}\}\), the three following functions are equal: \[ \sum_mf_{k,m}^{(N)}(\tau)\exp(2\pi imz), \] \begin{multline*} \frac{\left(a_{0}^{(p)}(2,1)g_{2-k,n_0-1}^{(p)}(z)+a_{0}^{(p)}(2,-1)g_{2-k,n_0+1}^{(p)}(z)+g_{2-k,n_0+2}^{(p)}(z)\right)f_{k,-n_0}^{(p)}(\tau)}{f_{0,2}^{(p)}(z)-f_{0,2}^{(p)}(\tau)} \\ + \frac{ g_{2-k,n_0-1}^{(p)}(z)\left(a_{0}^{(p)}(2,-1)f_{k,-n_0+2}^{(p)}(\tau)+f_{k,-n_0+3}^{(p)}(\tau)\right) } {f_{0,2}^{(p)}(z)-f_{0,2}^{(p)}(\tau)} \end{multline*} and \begin{multline*} \frac{\left(b_{0}^{(p)}(2,1)f_{k,-n_0}^{(p)}(\tau)+b_{0}^{(p)}(2,-1)f_{k,-n_0+2}^{(p)}(\tau)+f_{k,-n_0+3}^{(p)}(\tau)\right)g_{2-k,n_0-1}^{(p)}(z)}{g_{0,2}^{(p)}(z)-g_{0,2}^{(p)}(\tau)} \\ + \frac{ f_{k,-n_0}^{(p)}(\tau)\left(b_{0}^{(p)}(2,-1)g_{2-k,n_0+1}^{(p)}(z)+g_{2-k,n_0+2}^{(p)}(z)\right) } {g_{0,2}^{(p)}(z)-g_{0,2}^{(p)}(\tau)}. \end{multline*}