Analogues of Jacobi's derivative formula (Q2634562)

This paper is concerned with degree 1 theta functions of the form \[ \theta\bigg[{\varepsilon \atop \varepsilon'}\bigg](\zeta,\tau)=\sum_{n=-\infty}^{\infty}\mathbf{e}\Big(\big(n+\frac{\varepsilon}{2}\big)^{2}\frac{\tau}{2}+\big(n+\frac{\varepsilon}{2}\big)\big(\zeta+\frac{\varepsilon'}{2}\big)\Big),\qquad \mathrm{with}\quad\mathbf{e}(z)=e^{2\pi iz}, \] and in particular the theta constants obtained by substituting \(\zeta=0\) and the derivatives with respect to \(\zeta\) at \(\zeta=0\). The classical theta functions, with integral characteristics, have a definite parity, hence of the theta constant and the theta derivative, precisely one may not vanish. Jacobi's derivative formula states that the (odd) theta derivative \(\theta'\big[{1 \atop 1}\big](0,\tau)\) equals \(-\pi\) times the product of the (even) theta constants \(\theta\big[{0 \atop 0}\big](0,\tau)\), \(\theta\big[{0 \atop 1}\big](0,\tau)\), and \(\theta\big[{1 \atop 0}\big](0,\tau)\). Using the Jacobi triple product formula (which plays a prominent role in this paper), the latter expression can be shown to be the same as \(-2\pi\eta^{3}(\tau)\), and to give the explicit Fourier expansion of \(\eta^{3}\) (which contains only odd square powers of \(\mathbf{e}\big(\frac{\tau}{8}\big)\), where the \(n^{2}\)th coefficient is \(\big(\frac{-1}{n}\big)n\), in terms the Legendre symbol). Here \(\eta\) is the Dedekind eta function. This paper aims at generalizing these equalities for theta derivatives of some rational characteristics. The theta functions have some general properties: Translating \(\zeta\) by 1 or by \(\tau\) multiplies it by a well-known expression, changing the characteristics by even integers just multiplies the function by a simple constant, and changing them by any real numbers is the same as changing \(\zeta\) by the appropriate expression, up to again a known multiplier. The parity follows from the fact that inverting the characteristics is the same as inverting \(\zeta\), and each theta function vanishes at precisely one (known) point at a fundamental parallelogram (modulo 1 and \(\tau\)). In addition, the author uses classical results about the number of presentations of a given number as the sum of two squares or as the sum of a square and twice a square to identify when certain Fourier expansions yield \(\theta\big[{0 \atop 0}\big]^{2}(0,\tau)\) or \(\theta\big[{0 \atop 0}\big](0,\tau)\theta\big[{0 \atop 0}\big](0,2\tau)\). The first result of this paper is identifying the theta derivative \(\theta'\big[{1 \atop 1/2}\big](0,\tau)\) as \(-\pi\theta\big[{1 \atop 1/2}\big](0,\tau)\theta\big[{0 \atop 0}\big]^{2}(0,\tau)\), and identifying the Fourier expansion in which the coefficient of \(\mathbf{e}\big(\frac{n^{2}\tau}{8}\big)\) with odd \(n\) is \(\big(\frac{-2}{n}\big)n\) as the eta quotient \(\frac{\eta^{9}(2\tau)}{\eta^{3}(\tau)\eta^{3}(4\tau)}\). The first step is carried out by evaluating the logarithmic derivative of the Jacobi triple product expansion of \(\theta\big[{1 \atop 1/2}\big](\zeta,\tau)\) at \(\zeta=0\) (and identifying \(\theta\big[{0 \atop 0}\big]^{2}(0,\tau)\)), and then identifying the Jacobi triple product expansion of the result as roughly this eta quotient. For the second result one uses similar arguments for the logarithmic derivatives of \(\theta\big[{1 \atop 1/4}\big](\zeta,\tau)\) and \(\theta\big[{1 \atop 3/4}\big](\zeta,\tau)\) at \(\zeta=0\), and determines the associated theta derivatives using the fact that the difference (resp. sum) of these logarithmic derivatives is identified as \(2\pi\theta\big[{0 \atop 0}\big]^{2}(0,4\tau)\) (resp. \(-2\sqrt{2}\pi\theta\big[{0 \atop 0}\big](0,2\tau)\theta\big[{0 \atop 0}\big](0,4\tau)\)). The author then goes on to obtaining additional expressions for these theta derivatives, using the fact that certain quotients can be determined as independent of \(\zeta\) since the function of \(\zeta\) that they represent is elliptic and holomorphic. The paper concludes with more theta constant identities, resulting from the fact that certain families of 4 functions of \(\zeta\) must lie in a a space that is known to be 2-dimensional, and hence substituting some values of \(\zeta\) yield homogenous linear equations that must have a non-trivial solution. The vanishing of the resulting determinants yield the latter theta constant identities. Reviewer's remark: We conclude by noting that the short length of the paper is a bit misleading, as in several places the steps involving using the general properties of theta functions (parity, change of characteristics, etc.) are presented as trivial, with no details at all. For example, the relation of the expression for \(\theta'\big[{1 \atop 1/2}\big](0,\tau)\) and the aforementioned eta quotient requires the parity of theta functions, a fact that is not mentioned in the paper. This omission of details occurs several times. But a reader with interest in theta functions, who is willing to put some effort and care into following the calculations, will certainly be able to appreciate the results of this paper.