A generalization of Rosenhain's normal form for hyperelliptic curves with an application (Q676758)

The paper under review and the following one [see \textit{S. Koizumi}, ibid. 73, No. 1, 12-13 (1997; Zbl 0924.14017)] are devoted to the expression of cross-ratios of branching points of a hyper-elliptic curve in terms of \(\vartheta\)-constants. Although the principles of the derivation of such formulae are well known and well documented [see e.g. \textit{A. Krazer} and \textit{W. Wirtinger}, ``Abelsche Funktionen und allgemeine Thetafunctionen'', Encykl. Math. Wiss. II B 7, 604-873 (1920; JFM 47.0362.04), and \textit{H. M. Farkas} and \textit{I. Kra}, ``Riemann surfaces'' (1980; Zbl 0475.30001) for a modern exposition], the formula derived is explicit and useful for application. Consider a nonsingular hyperelliptic curve of genus \(g\) with branching points \(a_1,\dots,a_{2g+1},\infty\) with period matrix \(\Omega\). The associated \(\vartheta\)-function with characteristic \([\alpha]\) is defined as \[ \vartheta [\alpha](\Omega,w)= \sum_{l\in \mathbb{Z}^\sigma}\exp 2\pi\iota \{{1\over 2}\langle \ell+\alpha', (\ell+\alpha')\Omega \rangle+ \langle\ell+\alpha', w+\alpha'' \rangle\}, \] where \(w\in\mathbb{C}^g\), \(\alpha= (\alpha',\alpha'')\) are row vectors with \(\alpha',\alpha'' \in\mathbb{R}^g\) and \(\langle x,y\rangle=x \cdot{^ty}\). -- Put \(B=\{1,2,3, \dots,2g+1\}\), \(U=\{1,3,5, \dots,2g+1\}\). Define half-integer \(\vartheta\)-characteristics, \(\eta_k= (\eta_k', \eta_k'')\) by \[ \eta_{2i-1}'= \left(0,\dots, 0,{{\overset {i} {\breve 1}} \over 2},0,\dots, \right),\quad \eta_{2i-1}''=\left({1\over 2},\dots,{1\over 2}, {\overset {i}{\breve 0}}, 0,\dots,0\right), \] \[ \eta_{2i}'= \left(0,\dots, 0, {\overset {i}{\breve 1}\over 2},0,\dots,0\right), \quad\eta_{2i}'' =\left({1\over 2}, \dots,{1\over 2}, {\overset {i}{\breve 1}\over 2},0,\dots,0 \right), \] \(\eta_{2g+1}'= (0,\dots)\). For any subset \(T\) of \(B\) put \(\eta_T= \sum_{k\in T}\eta_k\) and for any subsets \(S,T\) denote by \(S\circ T=S\cup T-S\cap T\) and write \(\vartheta[T]= \vartheta[\eta_T] (\Omega,0)\). Theorem [Takase]. For any disjoint decomposition \(B=V\sqcup W\sqcup \{k,l,m\}\) with \(\#V= \#W=g-1\) we have \[ {a_k-a_l \over a_k-a_m}=\varepsilon (k;l,m) \left({\vartheta [U\circ (V \cup \{k,l\}] \vartheta[U \circ(W\cup \{k,l\}]\over \vartheta[U\circ (V\cup \{k, m\}] \vartheta[U\circ (W\cup \{k,m\}]} \right)^2. \] Here \(\varepsilon(k; l,m)= \begin{cases} 1\quad &\text{ if }k<l,m \text{ or }l,m <k,\\ -1\quad &\text{ if } l<k<m \text{ or }m< k<l.\end{cases}\) The theorem is proved in the paper under review on the basis of Frobenius addition theorem and Thomae formula. It is shown in the second paper (see the following review), that the result follows immediately from the classical formula of hyperelliptic dissection.