Before I state my question, let me give the set-up / "what I know".

Let $F / K$ be an extension of number fields of degree $n$. Let $v$ be a (discrete) valuation on $K$ and let $$ \mathfrak{O}_{v} = \{ \alpha \in K : v(\alpha) \geq 0 \} $$ be the valuation ring of $v$. Up to associates, $\mathfrak{O}_{v}$ has a unique irreducible element $p$ and $v(p) = 1$.

Let $v_1, \ldots, v_m$ be the extensions of $v$ to the field $F$, let $$ \mathfrak{O}_{v_i} = \{ \alpha \in F : v_i(\alpha) \geq 0 \} $$ be the valuation ring of $v_i$ $(i = 1, \ldots, m)$, and let $\pi_i$ be the unique (up to associates) irreducible element of $\mathfrak{O}_{v_i}$ $(i = 1, \ldots, m)$. Then the integral closure of $\mathfrak{O}_{v}$ in $F$ is $$ \overline{\mathfrak{O}}_{v} = \cap_{i=1}^{m} \mathfrak{O}_{v_i}, $$ the elements $\pi_1, \ldots, p_m$ constitute all the nonassociate irreducibles in $\overline{\mathfrak{O}}_{v}$, and $p$ factors in $\overline{\mathfrak{O}}_{v}$ as $$ p = \epsilon \pi_{1}^{e_1} \cdots \pi_{m}^{e_m} $$ where $\epsilon$ is a unit of $\overline{\mathfrak{O}}_{v}$, and where $e_i$ is the ramification index of $v_i$ with respect to $v$ (i.e., $e_i$ is the positive integer such that $v(\alpha) = v_i(\alpha)/e_i$ for all $\alpha \in K^{\times}$, i.e., $e_i = v_i(p)$).

The residue class field of $v$ is $\Sigma_{v} = \mathfrak{O}_v / m_v$, where $$ m_v = (p) = \{ \alpha \in K : v(\alpha) > 0 \} $$ is the unique maximal ideal of $\mathfrak{O}_v$. Similarly, the residue class field of $v_i$ is $\Sigma_{v_i} = \mathfrak{O}_{v_i} / m_{v_i}$, where $$ m_{v_i} = (\pi_i) = \{ \alpha \in F : v_i(\alpha) > 0 \} $$ is the unique maximal ideal of $\mathfrak{O}_{v_i}$. The residual degree of $v_i$ over $v$ is $f_i = [ \Sigma_{v_i} : \Sigma_{v} ]$

The following formula relating the ramification indices $e_i$, the residual degrees $f_i$, the number $m$ of extensions of $v$ to $F$, and the degree $n$ of $F / K$ holds: $$ \sum_{i=1}^{m} e_i f_i = n $$

Now the question: Suppose that $p$ factors in $\overline{\mathfrak{O}}_{v}$ as $$ p = \epsilon \pi^n $$ where $\epsilon$ is unit in $\overline{\mathfrak{O}}_{v}$ and $\pi$ is an irreducible element in $\overline{\mathfrak{O}}_{v}$.

Part 1: I want to show that $F = K(\pi)$ (that is, $\pi$ is a primitive element for $F/K$). To do this, it suffices to show the minimal polynomial of $\pi$ over $K$, $$ \phi(X) = X^k + a_{k-1}X^{k-1} + \cdots + a_0 \qquad (a_i \in K), $$ has degree $n$. Of course, $k \leq n$.

Alex Bartel has offered a solution below, but I don't understand some of the details. Here is my understanding of his argument. By definition, $$ 0 = \phi(\pi) = \pi^k + a_{k-1}\pi^{k-1} + \cdots + a_0 $$ On the left, for every $i$, $v_i(\phi(\pi)) = v_i(0) = \infty$. On the right, $v_i(\pi^k) = k$, $v_i(a_{j}\pi^{j}) = v_i(a_{j}) + j = nv(a_j) + j$ for $j=0, \ldots, k-1$, and so $$ v_i(\phi(\pi)) \geq \min(v_i(\pi^k), v_i(a_{j}\pi^{j})) = \min(k, nv(a_j) + j) $$ where $j$ ranges from $0$ to $k-1$. If $k \neq nv(a_j) + j \neq nv(a_l) + l$ for all $j \neq l$, then equality holds in the last displayed equation, which contradicts that the left-hand side is infinite. Alex asserts that $k \neq nv(a_j) + j \neq nv(a_l) + l$ for all $j \neq l$ holds if $k < n$, meaning $k < n$ is impossible. I do not understand why this assertion should be ture

Part 2: I also want to show that the minimal polynomial of $\pi$ over $K$, $$ \phi(X) = X^n + a_{n-1}X^{n-1} + \cdots + a_0 \qquad (a_i \in K), $$ is Eisenstein with respect to $v$ (that is, $v(a_{i}) > 0$ for $i = 1, \ldots, n-1$ and $v(a_0) = 1$ [I think this is what Eisenstein with respect to $v$ means]).

Based on the formula $\sum_{i=1}^{m} e_i f_i = n$, I know that $v$ has only one extension to $F$, $v_1$, with $e_1 = n$, $f_1 = 1$, and irreducible element $\pi_1 = \pi$. So $\pi$ is the unique irreducible element of $\overline{\mathfrak{O}}_{v}$ and every nonzero element $\alpha$ in $\overline{\mathfrak{O}}_{v}$ factors as $$ \alpha = \epsilon \pi^{a} $$ for some unit $\epsilon$ in $\overline{\mathfrak{O}}_{v}$ and some $a \geq 0$. Since the quotient field of $\overline{\mathfrak{O}}_{v}$ is $F$, any nonzero $\alpha$ in $F$ can be expressed as above if we allow $a$ to range through the integers.

That is where I am stuck.

I would like to solve this problem in the language of valuations as I am studying from Borevich and Shafarevich, and this is the language they use.

I would appreciate some help with this.


First, to show that $F=K(\pi)$, it is enough to show that $K(\pi)$ has degree $n$ over $K$, since one inclusion is clear. In fact, you only need to show that $K(\pi)$ has degree at least $n$. Let $v$ be the normalised valuation on $K$ corresponding to $p$ (which by the way means that $v(p)=1$, not 0 as you wrote), and extend $v$ to $F$, so that $v(\pi)=1/n$. Let $f=\sum_{i=1}^m a_mX^m$ be the minimal polynomial of $\pi$ over $K$. If $m<n$, then the valuations of all the monomials in $f(\pi)$ are distinct, since they are distinct modulo $\mathbb{Z}$, so $f(\pi)\neq 0$. You deduce that we must have $m=n$ and this settles your first question.

To show that $f$ is Eisenstein, you again consider valuations of the monomials. As we have already observed, if some monomial in $f(\pi)$ has valuation strictly smaller than all the others, then $f(\pi)\neq 0$. Since $v(\pi^n) = 1$, you deduce that $v(a_i)\geq 1$ for all $i$. But if $v(a_0)>1$, then the top power term has the lowest valuation. which we agreed couldn't be the case, so you are done.

  • $\begingroup$ I don't understand why the monomials of $f(\pi)$ have distinct valuations. What do you mean by modulo $\mathbb{Z}$? $\endgroup$ – admchrch Apr 1 '11 at 6:30
  • $\begingroup$ @admchrch The valuation of $\pi^i$ is $i/n$, the valuations of the coefficients $a_i$ are integers (since they are in the base field). So if $m < n$, then the valuations of the monomials have different fractional parts. That's what I mean by "modulo $\mathbb{Z}$". $\endgroup$ – Alex B. Apr 1 '11 at 11:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.