I have been trying to prove the following bijection which is a consequence of the nullstellansatz $$\{\text{maximal ideals of }\mathbb{C}[x_1,\dots,x_n] \} \leftrightarrow \{\text{points in }\mathbb{A}^{n}_{\mathbb{C}}\}$$

and I am struggling to prove that all maximal ideals in $\mathbb{C}[x_1,...x_n]$ have the form $\langle x_1 - a_1, .... , x_n - a_n\rangle $.

I have the following idea but it seems incorrect:

  1. Take a maximal ideal $M$ in $\mathbb{C}[x_1, \dots, x_n]:= C$.
  2. Then $C/ M$ is a field.
  3. Consider the ideals $\langle x_i + M \rangle$ in $C/M$.
  4. As these are ideals in a field they are either the zero ideal or the whole ideal.
  5. If they are all the zero ideal then we are done as then $\langle x_1,....,x_n \rangle$ is in $M$.
  6. Otherwise some $\langle x_i + M \rangle = C/M$ and by the correspondence theorem for rings when I pull this back to $C$ I get $\langle x_i \rangle = C$ which is a contradiction as $1 \not \in \langle x_i \rangle$.

I think this is wrong though as I don't think I have used algebraically closed anywhere. Thus I could use this to prove all maximal ideals in $\mathbb{R}[x]$ are of the forem $\langle x - a \rangle$ which is false as $\langle x^2 + 1 \rangle$ is maximal.

Is there a way to make this work?

  • 1
    $\begingroup$ $\langle x_i+M\rangle$ pulls back to $\langle x_i\rangle+M$ $\endgroup$ – anon May 2 '15 at 20:18

You had a proof that you were suspicious of, and you also had what looked like a counterexample to a statement that your proof looked like it proved. Did you try going through the steps of the proof with your counterexample? Here is what would have happened:

  1. Take a maximal ideal $(x^2 + 1)$ in $\mathbb{R}[x]$.
  2. Then $\mathbb{R}[x]/(x^2 + 1)$ is a field (it's $\mathbb{C}$).
  3. Consider the ideal $(x) + (x^2 + 1)$ in $\mathbb{C}$ (it's the ideal generated by $i$, which is all of $\mathbb{C}$).
  4. As this is an ideal in a field it is either the zero ideal or the whole field (it's the whole field).
  5. When I pull this back to $\mathbb{R}[x]$ I don't get $(x) = \mathbb{R}[x]$. As anon's comment says, I get $(x) + (x^2 + 1) = \mathbb{R}[x]$.

Anyway, this is not a statement that admits a completely formal proof. You really have to put in some amount of actual work. The key point is that $\mathbb{C}[x_1, x_2, \dots x_n]$ modulo a maximal ideal is always $\mathbb{C}$ (rather than some bigger field, as above). There are various ways to show this; one relatively cheap way (which depends on $\mathbb{C}$ being uncountable as well as algebraically closed) is to show that $\mathbb{C}[x_1, x_2, \dots x_n]$ modulo a maximal ideal cannot contain a transcendental element (over $\mathbb{C}$) because it's too small (at most countable-dimensional).


As pointed out above the method I posted is incorrect.

Here is a way to prove it using the weak nullstellansatz which states that: $$J \subsetneq \mathbb{C}[x_1, \dots, x_n] \implies V(J) \neq \emptyset.$$

Suppose $J$ is maximal. By definition $J$ is not the whole ring and so by the weak nullstellansatz $V(J) \neq \emptyset$.

Now notice that $J \subseteq I(V(J))$ and so $I(V(J))$ is either the entire ring or $J$. As $V(J)$ is not empty $I(V(J)) \neq \mathbb{C}[x_1, \dots, x_n]$ and so we are left with $J=I(V(J))$.

Now to see the form of $J$ is what we think it should be let $(a_1, \dots, a_n) \in V(J)$. Clearly $P:=\{(a_1,\dots,a_n)\} \subseteq V(J)$ and by the inclusion reversing of $I(\cdot)$ we see, $$J=I(V(J)) \subseteq I(P),$$

Using the maximality of $J$ again we get what we want, $$J = I(P) =\langle x_1 - a_1, \dots, x_n - a_n \rangle.$$


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.