Primitive divisors of elliptic divisibility sequences over function fields with constant \(j\)-invariant (Q2182155)

Let \((E,P)/F\) be a pair of an elliptic curve over a field \(F\) and a \(F\)-rational point \(P\) of \(E\) of infinite order. For a moment, let \(F=\mathbb Q\). For \(n\in\mathbb Z^+\), the \(x\)-coordinate of \(n\)-multiple point \([n] P\) of \(P\) is written in \(A_n/D_n^2, A_n,D_n\in\mathbb Z,(A_n,D_n)=1\). The sequence \(\{D_n\}_{n\in\mathbb Z^+}\) is called an elliptic divisibility sequence (EDS). \textit{J. H. Silverman} [Manuscr. Math. 114, No. 4, 431--446 (2004; Zbl 1128.11015)] showed that for almost all \(n\) \(D_n\) have a primitive valuation (divisor) \(v_p\), that is, \(v_p\) is a valuation associated with a prime number \(p\) such that \(v_p(D_n)>0\) and \(v_p(D_m)=0\) for all \(m<n\). If \(E\) has good reduction at \(p\), then \(v_p\) is a primitive valuation of \(D_n\) if and only if the reduction of \(P\) modulo \(p\) is a point of order \(n\). A positive integer \(N\) is a Zsigmondy bound of \(\{D_n\}_n\) if \(D_n\) has a primitive valuation for all \(n>N\). It is an unsolved problem whether there is a uniform Zsigmondy bound for all \((E,P)/\mathbb Q\). In the case that \(F\) is a number field, for the primitive valuations of EDS of a pair \((E,P)/F\) Silverman's result is also valid. Henceforth let \(F\) be a function field \(K(C)\) of a smooth, projective,geometrically irreducible curve \(C\) over a field \(K\). Then the EDS of a pair \((E,P)/F\) is defined as a sequence of divisors \(D_n=\sum_v m_v\cdot P_v\) on \(C\), where \(v\) runs over all valuations, \(P_v\) is the point on \(C\) corresponding to the valuation \(v\), \(x_v\) is the \(x\)-coordinate function for \(v\)-minimal Weierstrass equation of \(E\) and \(m_v=\max\{-v(x_v([n] P)),0\}/2\). If \(F\) is of characteristic \(0\), then the same result is true. In [New York J. Math. 22, 989--1020 (2016; Zbl 1417.11110)], the first author extended this to ordinary elliptic curves \(E\) over function fields of characteristic \(\ne 2,3\) and gave some estimates for Zsigmondy bound. All previous Zsigmondy bound estimates exclude the case of supersingular curves. Here, the authors consider Zsigmondy bounds of the EDS of a pair \((E,P)/F\) in the case that the \(j\)-invariant \(j(E)\in K\) and \((E,P)/F\) is not constant. A pair \((E,P)/F\) is said to be ``constant'' if there exists an elliptic curve \(E_0\) defined over \(K\) and an isomorphism \(\phi\) of \(E\) to \(E_0\times_K F\) defined over \(F\) such that \(\phi(P)\in E_0(K)\). Their results include the supersingular case. Principal results are as follows: (1) In the ordinary case, \(N=2\) is an optimal Zsigmondy bound, (2) In the supersingular case in characteristic \(p\), the terms \(D_n\) for \(n>8p\) have a primitive valuation if and only if \(p\nmid n\), and further a sharp version is obtained for every characteristic. Especially, suppose that \(E\) is defined over \(K\) and \(P\in E(F)\backslash E(K)\). Then (3a) if \(p\nmid n\) or \(E\) is ordinary, then \(D_n\) has a primitive valuation, (3b) if \(p|n\) and \(E\) is supersingular, then \(D_n=p^2D_{n/p}\) has no primitive valuation. The main idea of their proof is to compare EDS of \((E,P)\) with that of EDS of \((E_0,\phi(P))\) known by the results (3a) and (3b).