1
$\begingroup$

Question. Let $L:K$ be a field extension. Let $p, q, r\in K[x_1, \ldots, x_n]$, and suppose that there are polynomials $A, B\in L[x_1, \ldots, x_n]$ such that $pA+qB=r$. Do there necessarily exist polynomials $a, b\in K[x_1, \ldots, x_n]$ such that $pa+qb=r$?

In other words, if $I$ denotes the ideal generated by $p$ and $q$ in $K[x_1, \ldots, x_n]$, and $J$ denotes the ideal generated by $p$ and $q$ in $L[x_1, \ldots, x_n]$, then is it true that $I=J\cap K[x_1, \ldots, x_n]$?

I tried using induction on the sum of degrees of $p$ and $q$ but couldn't make it work.

1 Answers 1

0

We prove a general statement in the form of Lemma 2 below which easily implies the above speculated fact.

Lemma 1. Let $K:F$ be a field extension, giving a map $F[x_1 , \ldots, x_n]\to K[x_1 , \ldots, x_n]$. If $\{v_\alpha\}_{\alpha\in J}$ is an $F$-basis for $K$, then we have $$K[x_1 , \ldots, x_n]= \bigoplus_{\alpha\in J} F[x_1 , \ldots, x_n]v_\alpha$$ as $F[x_1 , \ldots, x_n]$-modules.

We write $\mathbf x$ to denote the list $x_1 , \ldots, x_n$. It is clear that $\sum_{\alpha\in J}F[\mathbf x]v_\alpha= K[\mathbf x]$. So let $p_1 , \ldots, p_k\in F[\mathbf x]$ be such that $p_1 v_{\alpha_1}+ \cdots + p_kv_{\alpha_k}=0$. We want to show that each $p_i$ is $0$ to finish the proof. Let $m:=x_1^{a_1} \cdots x_n^{a_n}$ be a monomial. Let the coefficient of $m$ in $p_i$ be written as $c_i$, which is a member of $F$. Then the coefficient of $m$ in $p_1v_{\alpha_1}+ \cdots + p_kv_{\alpha_k}$ is $c_1v_{\alpha_1}+ \cdots + c_kv_{\alpha_k}$. But this is also equal to $0$. Since $\{v_\alpha\}_{\alpha\in J}$ form an $F$-basis for $K$, we have each $c_i$ is $0$. Therefore, the coefficient of any monomial in each $p_i$ is $0$, meaning that each $p_i$ is $0$, and we are done.

Lemma 2. Let $K:F$ be a field extension, giving a map $F[x_1 , \ldots, x_n]\to K[x_1 , \ldots, x_n]$. Let $I$ be an ideal in $F[x_1 , \ldots, x_n]$. Then $I^{ec}=I$.

We write $\mathbf x$ to denote the list $x_1 , \ldots, x_n$. Since $F[\mathbf x]$ is Noetherian, there are finitely many polynomials $p_1 , \ldots, p_k\in F[\mathbf x]$ such that $I=\langle {p_1 , \ldots, p_k}\rangle$. Let $p\in I^{ec}$. Then there are polynomials $f_1 , \ldots, f_k\in K[\mathbf x]$ such that $p=f_1 p_1 + \cdots + f_kp_k$. Let $\{v_\alpha\}_{\alpha\in J}$ be an $F$-basis for $K$, where, say $v_{\alpha_0}=1\in F$. By Lemma 1, each $f_i$ can be written as

$$ f_i = \sum_{\alpha\in J} v_\alpha f_i^\alpha$$ where all but finitely many $f_i^\alpha$ are $0$ and each $f_i^\alpha\in F[\mathbf x]$. Thus we have $$p = \sum_{\alpha\in J} v_\alpha(f_1^\alpha p_1 + \cdots + f_k^\alpha p_k)$$ which gives $$p = (f_1^{\alpha_0}p_1+ \cdots + f_k^{\alpha_0}p_k) + {\sum_{\alpha\in J\setminus\{\alpha_0\}} v_\alpha(f_1^\alpha p_1 + \cdots + f_k^\alpha p_k) }$$ But since $p\in F[\mathbf x]$, by Lemma 1, we must have

$$p= f_1^{\alpha_0} p_1 + \cdots + f^{\alpha_0}_k p_k$$ and thus $p\in I$. So we have shown that $I^{ec}\subseteq I$. The reverse containment is clear, and thus we have $I^{ec}=I$.