Reduction and specialization of hyperelliptic continued fractions
Abstract
For a monic polynomial D(X) of even degree, express D as a Laurent series in X-1; this yields a continued fraction expansion (similar to continued fractions of real numbers): \[ D=a0+1a1+1a2+1, ai polynomials in X.\] Such continued fractions were first considered by Abel in 1826, and later by Chebyshev. It turns out they are rarely periodic unless D is defined over a finite field. Around 2001 van der Poorten studied non-periodic continued fractions of D, with D defined over the rationals, and simultaneously the continued fraction of D modulo a suitable prime p; the latter continued fraction is automatically periodic. He found that one recovers all the convergents (rational function approximations to D obtained by cutting off the continued fraction) of D p by appropriately normalising and then reducing the convergents of D. By developing a general specialization theory for continued fractions of Laurent series, I produced a rigorous proof of this result stated by van der Poorten and further was able to show the following: If D is defined over the rationals and the continued fraction of D is non-periodic, then for all but finitely many primes p ∈ Z, this prime p occurs in the denominator of the leading coefficient of infinitely many ai. For deg\,D = 4, I can even give a description of the orders in which the prime appears, and the p-adic Gauss norms of the ai and the convergents. These results also generalise to number fields. Moreover, I derive optimised formulae for computing quadratic continued fractions, along with several example expansions. I discuss a few known results on the heights of the convergents, and explain some relations with the reduction of hyperelliptic curves and Jacobians.