Key polynomials and pseudo-convergent sequences (Q1682073)

Let \(v\) be a valuation on a field \(K\). In order to describe all the possible extensions of \(v\) to the polynomial ring \(K[X]\) (or, equivalently, to its quotient field \(K(X)\)), \textit{S. MacLane} developed in [Trans. Am. Math. Soc. 40, 363--395 (1936; Zbl 0015.29202)] the theory of key polynomials. If \(v\) is discrete, MacLane's work describes every possible extension as a limit of a sequence of augmented valuations defined by means of key polynomials. This work has been generalized by \textit{M. Vaquié} for a general valuation \(v\) in [Trans. Am. Math. Soc. 359, No. 7, 3439--3481 (2007; Zbl 1121.13006)]. Another approach was given by \textit{A. Ostrowski} a few years before MacLane in [Math. Z. 39, 269--404 (1934; Zbl 0010.15001)]. He introduced the notion of pseudo-convergent sequence of a rank one valued field \(K\). Given an extension \(u\) of \(v\) to the algebraic closure \(\overline{K}\), then every possible rank one extension of \(v\) to \(K(X)\) is determined by a pseudo-convergent sequence contained in \(\overline{K}\) (with respect to \(u\)). In this article under review, the following new notion of key polynomial is introduced: for a positive integer \(n\) let \(\partial_n=\frac{1}{n!}\frac{\partial^n}{\partial x^n}\) be the operator given by the \(n\)-th formal derivative. Let \(\nu\) be a valuation on \(K[X]\). Given a polynomial \(f\in K[X]\), we set \(\epsilon(f)=\max_{n\in\mathbb N}\{\frac{\nu(f)-\nu(\partial_n f)}{n}\}\). A monic polynomial \(Q\in K[X]\) is said to be a \textit{key polynomial of level} \(\epsilon(Q)\) if for every \(f\in K[X]\), whenever \(\epsilon(f)\geq \epsilon(Q)\), then \(\deg(f)\geq \deg(Q)\). In the first part of the paper, the authors start with a valuation \(\nu\) on \(K[X]\) and, given a monic polynomial \(q\in K[X]\), define the \(q\)-truncation of \(\nu\) as the map defined by \(\nu_q(f)=\min_i\{\nu(f_iq^i)\}\), for every \(f\in K[X]\), with \(f=\sum_i f_iq^i\) being the \(q\)-standard expansion of \(f(X)\) (i.e., \(f_i\in K[X]\) with either \(f_i=0\) or \(\deg(f_i)<\deg(q)\)). The authors prove that the \(q\)-truncation of \(\nu\) associated to a key polynomial is a valuation (Proposition 2.6) and also shows that in general \(\nu_q\) is not a valuation (Example 2.5). Furthermore, the main result Theorem 1.1 of this part shows that every valuation \(v\) on \(K[X]\) admits a complete set of key polynomials \(\Lambda\), in the sense that for each \(f\in K[X]\) there exists \(Q\in \Lambda\) such that \(\nu(f)=\nu_Q(f)\). Theorem 2.12 gives a criterion for a monic polynomial \(Q\in K[X]\) to be a key polynomial. In the second part, the authors determine a complete set of key polynomials for a valuation \(\nu\) on \(K[X]\) for which \(X\) is a pseudo-limit of a pseudo-convergent sequence \(\{a_{\rho}\}_{\rho<\lambda}\) in \(K\), either in the case \(\{a_{\rho}\}_{\rho<\lambda}\) is of transcendental type or is of algebraic type.