Indefinite theta series of signature \((1,1)\) from the point of view of homological mirror symmetry (Q2566192)

Let \(\Lambda\) be a rank \(2\) lattice equipped with a non-degenerate \({\mathbb Q}\)-valued quadratic form \(Q\) of signature \((1,1)\). Fix an open cone \(C\subset\Lambda_{\mathbb R}\) of the form \(C=\{{\mathbf v}\in\Lambda_{\mathbb R}\mid \phi({\mathbf v})\cdot\psi({\mathbf v})>0\}\) for a pair of linear forms \(\phi\) and \(\psi\) on \(\Lambda_{\mathbb R}\) such that \(Q| _C>0\). Let \(C=C^+\cup C^-\) be the decomposition of \(C\) into connected components, \(\text{sign}: C\to\{\pm 1\}\) be the sign function such that \(\text{sign}(C^{\pm 1})=\pm 1\), respectively. Let \(\alpha:\Lambda_{\mathbb C}\to \Lambda_{\mathbb R}\) be the map given by \(\alpha({\mathbf z})= \text{Im}({\mathbf z})/ \text{Im}(\tau)\) where \(\tau\) belongs to the upper-half plane \({\mathcal H}\). Definition: The indefinite theta series associated with \((\Lambda, Q, C)\) is defined by \[ \Theta_{\Lambda,Q,C}({\mathbf z},\tau) = \sum_{{\mathbf n}\in\Lambda, {\mathbf n}+\alpha({\mathbf z})\in C} \text{sign}({\mathbf n}+\alpha({\mathbf z})) \exp(\pi i\tau Q({\mathbf n})+2\pi i{\mathbf n}\cdot{\mathbf z}). \] For some special choices of \(\Lambda, Q, C\), vectors \({\mathbf v}, {\mathbf w}\in\Lambda_{\mathbb Q}\) and a rational number \(\lambda\), it is known that the function \(\exp(\pi i\lambda)\Theta_{\Lambda,Q,C}(\tau{\mathbf v}+{\mathbf w},\tau)\) is a modular form of weight \(1\) for some congruence subgroup of SL\(_2({\mathbb Z})\). The point of this article is to provide examples in support of the following conjecture. Conjecture: An indefinite theta series of signature \((1,1)\) is modular if and only if it corresponds to a universal univalued triple Massey product in the derived categories of coherent sheaves on elliptic curves. Some of the results are formulated as follows. Theorem 1: Every indefinite theta series of signature \((1,1)\) is expressed in terms of theta series associated with split quadratic forms and the usual theta series \(\theta_c\). More concretely, given a triple \((\Lambda, Q, C)\) and a connected open subset \(U\subset \Lambda_{\mathbb C}\setminus \alpha^{-1}(\partial C+\Lambda)\), there exist \({\mathbb Q}\)-linear functionals \((r, s, \ell_i\), \(i=1,\dots, N)\) on \(\Lambda_{\mathbb Q}\), constants \((a_i, b_i, c_i, d_i, e_i\), \(i=1,\dots, N)\) in \({\mathbb Q}\), \(f\in{\mathbb Q}_{>0}\) and meromorphic elliptic functions \((\phi_i, \psi_i\), \(i=1,\dots, N)\) such that \[ \Theta_{\Lambda, Q, C}({\mathbf z},\tau) =\sum_{i=1}^N\phi_i(r({\mathbf z}))\psi_i(s({\mathbf z}))\kappa_{e_i}(a_i\tau+b_i, \ell_i({\mathbf z}) +c_i\tau+d_i; f\tau) \] for \({\mathbf z}\in U\), \(\tau\in\mathcal H\). Here a meromorphic function \(\phi(z,\tau)\) on \({\mathbb C}\times {\mathcal H}\) is said to be elliptic if it can be expressed rationally over \({\mathbb C}\) in terms of functions of the form \(\theta_c(az+b\tau, d\tau)\) and \(\exp(\pi id\tau)\), where \(a,b,b\in{\mathbb Q}\), \(d\in{\mathbb Q}_{>0}\) and \(\theta_c(z,\tau) =\sum_{n\in{\mathbb Z}+c} \exp(\pi i\tau n^2+2\pi inz)\). Also we denote \[ \kappa_c(y,x;\tau)=\sum_{m\in{\mathbb Z}+c} {{\exp(\pi i \tau m^2+2\pi imx)}\over{\exp(2\pi im\tau)- exp(2\pi iy)}}. \] Proof of Theorem 1 is based on homological mirror symmetry for elliptic curves, and uses the interpretation of indefinite theta series as components of triple Fukaya products on a symplectic torus and the \(A_{\infty}\)-category connecting \(m_2\) and \(m_3\). The modularity conjecture of indefinite theta series is addressed in the following theorem. Theorem 2: Indefinite theta series corresponding to univalued Massey products between line bundles on elliptic curve are modular. As corollaries, interesting new identities between \(q\)-series are obtained. For instance, \[ \sum_{(m,n)\in{\mathbb Z}^2,m\equiv n+1 (2)}^{\text{indef}} (-1)^{(m+n-1)/2} q^{(m^2+6mn+3n^2)/2} ={{\eta^3(2\tau)\theta_{1/2}(1/4,3\tau)}\over{\theta(1/2,4\tau)\theta_{1/2}(1/4,\tau)}} \] \[ =q^{1/2}\prod_{n\geq 1}(1+q^n)(1-q^{2n})(1-q^{3n})(1+q^{6n}). \] Here \(\eta(\tau)\) is the Dedekind eta-function.