The defect formula (Q6110341)

In this paper a valuation \(\mu\) is not assume to satisfies \(\mu (x)=\infty\Leftrightarrow x=0\). The set supp\((\mu)=\mu^{-1}(\{\infty\})\) is a prime ideal which is called the support of \(\mu\). What is commonly called valuation is here called valuation with trivial support. In the following \(K\) is a field, \(K[x]\) is the ring of polynomials. The second result of this paper deals with the graded algebra associated to a valuation \(\mu\) on \(K[x]\). For every \(\alpha\in \mu(K[x])\), let \(\mathcal{P}_{\alpha}=\{f\in K[x]: \mu(f)\geq \alpha\}\), \(\mathcal{P}_{\alpha}^+=\{f\in K[x] : \mu(f)> \alpha\}\). The graded algebra of \(\mu\) is \(\displaystyle{\mathcal{G}_{\mu}=\oplus_{\alpha\in \mu(K[x])} \mathcal{P}_{\alpha}/\mathcal{P}_{\alpha}^+ }\). For every \(f\in K[x]\), we let in\(_{\mu}(f)\) be its image in \(\mathcal{P}_{\mu(f)}/\mathcal{P}_{\mu(f)}^+\). Denote by \(K^h\) the Hensel closure of \((K,\mu)\), and by \(\mu^h\) the extension of \(\mu\) to \(K^h[x]\). Then \(\mathcal{G}_{\mu}\) embeds in a natural way in \(\mathcal{G}_{\mu^h}\). The second theorem of this paper states that this embedding is an isomorphism. Now we require more definitions. Let \(\phi\) be a key polynomial for \(\mu\), \(\gamma>\mu(\phi)\) and \(a_0,a_1,\dots,a_n\) in \(K[x]\) with degrees less than the degree of \(\phi\). We set \(\nu (a_0+a_1\phi+\cdots +a_n\phi^n)=\min \{\mu(a_0),\mu(a_1)+\gamma,\dots,\mu(a_n)+n\gamma\}\). The mapping \(\nu\) is called an augmentation of \(\mu\). We define \(\rho_{\phi}\) in the same way, by taking \(\mu(\phi)\) instead of \(\gamma\). Now, let \((\nu_i)_{i\in I}\) be a sequence of augmentations of \(\mu\) as before, where all the key polynomials have the same degree and \(\nu_{i+1}\) is also an augmentation of \(\nu_i\) as before, and such that for every \(f\in K[x]\) the sequence \((\nu_i(f))_{i\in I}\) is eventually constant, and denote by \(\nu(f)\) this limit. Then \(\nu\) is also called an augmentation of \(\mu\). If \(\nu\) is an augmentation, then for every \(f\in K[x]\) we have \(\mu(f)\leq \nu(f)\); this is denoted by \(\mu<\nu\) in the case where there is some \(f\) such that the inequality is strict. If \(\nu\) is an augmentation of \(\mu\), then \(t(\mu,\nu)\) denotes the set of all monic \(Q\in K[x]\) of minimal degree such that \(\mu(Q)<\nu(Q)\). Now, let \(\phi\in K[x]\) be a monic polynomial such that either supp\((\nu)=\phi K[x]\) or \(\phi\) is a key polynomial of minimal degree of \(\nu\). Then the minimum of the set \(\{\deg(in_{\rho_Q}(\phi) : Q\in t(\mu,\nu)\}\) is independent of the choice of \(\phi\). This minimum is called the defect of the augmentation, and it is denoted by \(d(\mu \rightarrow \nu)\). The first theorem of this paper asserts that \(d(\mu \rightarrow \nu)= d(\mu^h \rightarrow \nu^h)\), where \(\nu^h\) is the corresponding augmentation of \(\mu^h\). Now, we turn to a main ingredient of the proof of these theorems. Let \(v\) be another valuation on \(K\) such that \((K,v)\) is henselian, and we still denote by \(v\) its extension to the algebraic closure \(\overline{K}\) of \(K\). Let \(F\in K[x]\) be irreducible and \(\theta\) be a root of \(F\) in \(\overline{K}\). Then for every \(f\in K[x]\) the value \(v(f(\theta))\) is independent of the choice of the root \(\theta\). This value is denoted by \(v_F(f)\). Let \(\phi\) be a key polynomial for \(\mu\). Then for every irreducible \(F\in K[x]\), in\(_{\mu}(\phi)\) divides in\(_{\mu}(F)\) if, and only if, \(\mu<v_F\), and \(t(\mu,v_F)\) is equal to the set of all key polynomials \(\phi\) for \(\mu\) such that in\(_{\mu}(\phi)=\) in\(_{\mu}(\phi)\). Furthermore, if this holds, then in\(_{\mu}(F)=\) in\(_{\mu}(\phi^l)\), with \(l=\)deg\((F)/\)deg\((\phi)\), and either \(F=\phi\), or the Newton polygon of \(F\) is one slide of slope \(-v_F(\phi)\). This last condition means that, by letting \(F=a_0+\cdots +a_n\phi^n\), with \(a_0,\dots,a_n\) in \(K[x]\) with degrees less than the degree of \(\phi\), we have, for every \(j\in \{0,\dots, n\}\), \(\mu(a_n)\geq -v_F(\varphi) j+\mu(a_0)\), and equality holds for \(j=0\) and some \(j>0\).