# No maximal ideal of $k[x_1, \dots, k_n]$ is principal for $n > 1$ ($k$ any field)

This problem is from a qualifying exam.

I am to show that if $k$ is a field and if $n > 1$, then no maximal ideal of $k[x_1, \dots, x_n]$ is principal.

I understand why this is true if $k$ is algebraically closed, since every maximal ideal is of the form $(x_1 - a_1, \dots, x_n - a_n)$ for some $a_i \in k$.

My attempt for $k$ an arbitrary field. Let $m \subset k[x_1, \dots, x_n]$ be a maximal ideal and let $K$ be the algebraic closure of $k$. Let $M \subset K[x_1, \dots, x_n]$ be the push-forward ideal of $m$, namely, $M = K[x_1, \dots, x_n]m$. Assume $M$ is not the unit ideal, and so is contained in the maximal ideal $(x_1-a_1, \dots, x_n - a_n)$ of $K[x_1, \dots, x_n]$ for some $a_i \in K$.

Suppose $m$ is principal and generated by some irreducible $f \in k[x_1, \dots, x_n]$. Observe that the subset $A$ of $(x_1 - a_1, \dots, x_n - a_n)$ consisting of all polynomials whose coefficients are in $k$ is an ideal of $k[x_1, \dots, x_n]$ containing $m$, and therefore $A = m$ by maximality. For each $i$, let $f_i \in k[x_1, \dots, x_n]$ be a polynomial of minimal degree such that $x_i - a_i$ divides $f_i$ in $K[x_1, \dots, x_n]$ -- such a polynomial exists since each $a_i$ is algebraic over $k$. In particular, $f_i \in A = m$. Since $f$ divides $f_i$ in $k[x_1, \dots, x_n]$, it follows by the minimality of $f_i$ that $f_i = f$ for all $i$. We gather $x_i - a_i$ divides $f$ for all $i$.

Note that if any of the $a_i$ are in $k$ then we are done because $f = f_i = x_i - a_i$, again by the minimality of $f_i$, which is a contradiction since $x_j - a_j \nmid x_i - a_i$ for $i \neq j$.

However, I cannot find the contradiction in the case all the $a_i$ are in $K - k$. Does anyone see how to proceed along this particular route? If not, I welcome other suggestions. I understand there are arguments using the Krull dimension, but given that Krull dimension is not covered in our algebra course, I'm hesitant to use such an argument on an exam. I think it is intended we use Nullstellensatz, but I'm not entirely sure.

EDIT: As discussed in the comments, one also must account for the case $M$ is unital, if this strategy is to be kept.

• Why $M$ is not the unit ideal? thanks. – Matemáticos Chibchas Sep 5 '16 at 22:33
• Because $m$ is not, and we're just multiplying $m$ by polynomials to get $M$ – Doug Sep 5 '16 at 22:34
• This is not sufficient; when you talk about "polynomials" you are being ambiguous (are they with coefficients in $K$ or in $k$?). – Matemáticos Chibchas Sep 5 '16 at 22:59
• Multiplying $m$ by polynomials with coefficients in $K$. This cannot decrease the degree, so $1$ remains outside of $M$. – Doug Sep 5 '16 at 23:02
• Now I see... but note that you are already assuming that $m$ is principal at this point. In the general case (that is, $m$ not principal) your degree argument doesn't work. Only on the next paragraph you are assuming that $m$ is principal; you must correct that. – Matemáticos Chibchas Sep 5 '16 at 23:06

You're almost done: what you showed is that if $$f_i(x)$$ is the minimal polynomial of $$a_i$$ over $$k$$, then $$f_i(x_i) \in \mathfrak{m}$$ is divisible by $$f(x_1, ..., x_n)$$ in $$k[x_1, ..., x_n]$$, hence they are equal. Since $$n \geq 2$$ this contradicts that $$f$$ is irreducible. (In other words, it's not possible for all $$f_i$$ to be equal to $$f$$, because the $$f_i$$ involve different variables from each other.)

• What I can say for sure is that $f = (x_1 - a_1)...(x_n - a_n)g$ for some polynomial $g$ with coefficients in $K$. I don't see that this immediately contradicts the irreducibility of $f$ over $k$. I don't think we can say, a priori, that the $f_i$ involve different variables from each other. – Doug Aug 26 '16 at 4:29
• In characteristic zero you can say it like this: if $x_i - a_i$ divides $f$, a polynomial in $k[x_1, ..., x_n]$, as polynomials in $K[x_1, ..., x_n]$, then also all the galois conjugates $x_i- \sigma(a_i)$ of $x_i-a_i$ divide $f$, so the minimum polynomial of $a_i$ divides $f$. In other words, if $\mathbf{M} = (x_1-a_1, ..., x_n-a_n)$ contains $\mathbf{m} K[x_1, ..., x_n]$, then so do all the galois conjugates of $\mathbf{M}$. (If you talk about minimal polynomials directly then what you say makes sense in positive characteristic too.) – anon Aug 26 '16 at 4:56
• Ah, I think I understand what you are saying. If $m_i(X) \in k[X]$ is the minimal polynomial of $a_i$ over $k$, then $x_i-a_i$ divides $m_i(x_i)$ in $K[x_1, \dots, x_n]$, hence $m_i(x_i) \in m$ and $f$ divides $m_i(x_i)$ in $k[x_1, \dots, x_n]$. We want to conclude $f$ and $m_i(x_i)$ are associates for all $i$, a contradiction. Possibly a silly question: By assumption $m_i(x_i)$ is irreducible in $k[x_i]$; do we know for sure it remains irreducible in $k[x_1, \dots, x_n]$? – Doug Aug 26 '16 at 5:37
• Yes, that's it. For the irreducibility, $k[x_1, ..., x_n]$ is graded by degree in each of the variables $x_1,..., x_n$; that is, the degree of a product $fg$ in a given variable $x_j$ is the sum of the degrees of $f$ and $g$ in $x_j$. This implies that a polynomial in which only a proper subset of variables appears does not factor into a product where the other variables appear. – anon Aug 26 '16 at 16:07

Krull’s principal ideal theorem. If $$A$$ is Noetherian and $$f$$ is a non-unit, then any prime ideal $$P$$ that is minimal with respect to containing $$(f)$$ is necessarily of height at most one.

In particular, if $$(f)$$ is prime, it's of height at most one. Now make a Krull dimension argument to show that $$(f)$$ can't be maximal. Also, don't forget to justify why you're in a Noetherian ring.

• I am looking for an answer that doesn't use the Krull dimension, possibly using Nullstellensatz as I have attempted to do. I edited my question to state this explicitly. – Doug Aug 26 '16 at 2:40
• @user26857 The justification of why the ring is Noetherian is to verify that we can actually apply the theorem. I just mentioned that because it seemed like the asker wanted to be as thorough as possible. Also, a possible argument for the Krull dimension part is to say that since we're in an f.g. $k$-algebra, $\dim k[x_1, \dots, x_n]/(f) = \dim k[x_1, \dots, x_n] - height((f))$. In particular it's $n-1$. Now use the correspondence theorem to get that this is the size of a chain of primes in $k[x_1, \dots, x_n]$ containing $(f)$. Since $n \geq 2$, $(f)$ can't be maximal. – Samuel Yusim Aug 27 '16 at 5:04