Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to understand a certain scenario and to do so, I sat down and calculated an explicit example. While doing so, I was able to "prove" two statements that directly contradict each other, which is (hopefully) not possible. I would like to ask you to help me find my mistake.

Let $n>2$. Consider the polynomial ring $A=\mathbb{C}[x,y]$ and the element $a:=x+2$. Set

$B := \mathbb{C}[x,y,u,v,w]/(u^n-x,v^n-y,w^n-a)$

Classical Understanding: $B$ is the coordinate ring of the affine variety

       $Y=Z(u^n-x,v^n-y,w^n-x-2)\subset \mathbb{A}_{\mathbb{C}}^5$

and composing this inclusion with the projection $\mathbb{A}_{\mathbb{C}}^5\to\mathbb{A}_{\mathbb{C}}^2$ to the first two coordinates, we obtain a map $\pi:Y\to\mathbb{A}_{\mathbb{C}}^2$. For any $\zeta\in\sqrt[n]{2}\subset\mathbb{C}$, the point $(0,0,0,0,\zeta)$ is contained in $Y$, so there are at least $n$ preimages of the origin $(0,0)\in\mathbb{A}_{\mathbb{C}}^2$ under $\pi$. In fact, one can easily check that there are exactly $n$ preimages, but we shall not concern ourselves with this.

Modern Understanding: The canonical map $f:A\to B$ induces a morphism of varieties


which is given by pulling back prime ideals via $f$. Let $P$ be the maximal ideal of $A$ generated by $x$ and $y$, hence the one corresponding to the origin. Assume that $Q\subset B$ is a prime ideal such that $f^{-1}(Q)=P$, i.e. $Q\in\pi^{-1}(P)$. Let us denote the image of $a\in\mathbb{C}[x,y,u,v,w]$ in $B$ by $\overline{a}$.

Since $x\in P$ and $f(x)\in f(P)\subset Q$, we have $\overline{x},\overline{y}\in Q$. Since $\overline{u}^n=\overline{x}$ and $Q$ is prime (hence radical), we have $\overline{u},\overline{v}\in Q$ and we observe that $Q':=(\overline{u},\overline{v})$ already satisfies $f^{-1}(Q')=P$. Hence, by the Going-Up theorem for integral extensions (Eisenbud Proposition 9.2), we know that $Q'$ is maximal, hence $Q'=Q$. Alternatively, one can easily check that all elements of $B$ are mapped to elements of $\mathbb{C}$ under $p:B\twoheadrightarrow B/Q'$: This is clear for $\overline{u}$, $\overline{v}$, $\overline{x}$ and $\overline{y}$, and we have


so $p(\overline{w})$ is some $n$-th root of $2$.

Either way, the point $P$ has exactly one preimage under $\pi$, namely $Q=Q'$.

So, it is my understanding that I describe the same scenario in different languages, so the origin must have either one or $n$ preimages, but both can not be the case.

share|cite|improve this question
Minor quibble: $(0,0,\zeta,0,0)$ is not in $Y$, $(0,0,0,0,\zeta)$ is. – Craig Sep 8 '11 at 15:18
corrected, thanks. – Jesko Hüttenhain Sep 8 '11 at 15:27
up vote 4 down vote accepted

To begin with, note that the elements $u,v$ are not in $B$ so your definition of $Q'$ doesn't really make sense. I suppose you mean $Q'=(\bar u,\bar v)\subset B$, and I'll assume that.

The key to your paradox is quite simple: the ideal $Q'\subset B$ is not prime, and thus not maximal! Indeed if we consider the projection $q:\mathbb C[x,y,u,v,w]\stackrel {def}{=}R \to B$ we get $q^{-1}(Q')=(x,y,u,v,w^n-2)\subset R$ which is clearly not prime. Actually, if we call $Q(\zeta)\stackrel {def}{=}(\bar x,\bar y,\bar u,\bar v,\bar w-\zeta) \subset B $ , a maximal ideal, we have $Q'=\bigcap \limits_{\zeta^n=2} Q(\zeta)=\prod \limits _{\zeta^n=2} Q(\zeta)$, so that $V(Q')\subset Y$ consists of the $n$ reduced, rational points $Q(\zeta) \in Y$.
By the way, the schematic fiber $V(\bar x, \bar y)\subset Y$ of the origin of $\mathbb A^2_{\mathbb C}$ under $\pi $ is not reduced and has as reduction $V(Q')$.

Edit: Answer to question in comment. The surface $Y$ is indeed smooth (it is isomorphic to the product of the affine $v$-line with the plane curve $w^n-u^n-2=0$ in the $w,u$-plane).
Hence the ideal $Q(\zeta)$ can be described by two generators near the point $Q(\zeta)$ , namely
$\bar v$ and $\bar u$. If this seems strange remember that:

$\bar x=\bar u^n,\bar y=\bar v^n $ and $\bar w-\zeta=\frac {\bar u^n}{\bar w^{n-1}+\ldots+ \zeta^{n-1}}$ in the local ring of the point $Q(\zeta)\;$

[since $0=\bar w^n-\bar u^n-2=\bar w^n-\bar u^n- \zeta^n=(\bar w-\zeta)(\bar w^{n-1}+\ldots+ \zeta^{n-1})-\bar u^n \;$ ]

share|cite|improve this answer
Your assumption is completely correct (I fixed that in the question), as is your answer! Thanks, that has relieved me from a major headache. I have another question, however: I think that $Y$ is regular, so $Q(\zeta)$ should be locally generated by two elements. Can you tell me which? – Jesko Hüttenhain Sep 8 '11 at 18:38
Dear @rattle, I have answered your question in an edit. You seem to have a very active way of learning algebraic geometry by linking the language of classical varieties to that of schemes and making interesting explicit calculations. Congratulations! – Georges Elencwajg Sep 8 '11 at 20:50
Ah, let $\sqrt[n]{2}=\{\zeta_1,\ldots,\zeta_n\}$. Then, locally around $Q(\zeta_i)$, the $\overline{w}-\zeta_j$ for $j\ne i$ become units, so $\prod_{j\ne i} \frac{1}{\overline{w}-\zeta_j} \cdot \overline{u}^n = \overline{w}-\zeta_i$ and therefore, the maximal ideal $\mathfrak{m}_{Q(\zeta_i)}$ is, in fact, generated by the images of $\overline{v}$ and $\overline{u}$. How ironic. – Jesko Hüttenhain Sep 8 '11 at 20:55
I swear, I figured it out mere seconds before you posted. Thanks a bunch, though, now I finally understand said scenario. – Jesko Hüttenhain Sep 8 '11 at 20:57
Dear@rattle: I completely believe you! It has often happened that the moment I post some text, I see that somebody else has just posted a practically identical text a few minutes before, which gives the impression that I have just copied it! – Georges Elencwajg Sep 8 '11 at 21:15

I don't think we know that the map $p$ is uniquely defined.

share|cite|improve this answer
It's the projection to the quotient, I very much think it is. – Jesko Hüttenhain Sep 8 '11 at 15:30
You're right. Sorry. The problem is that the quotient is not $\mathbb{C}$, it's $\mathbb{C}[\hat{w}]/(\hat{w}^n -2)$, where $\hat{w}$ is the image of $w$ under the projection. – Craig Sep 8 '11 at 15:47
$\mathbb{C}[x]/(x^n-2)=\mathbb{C}$ since $\mathbb{C}$ is algebraically closed. – Jesko Hüttenhain Sep 8 '11 at 16:33
I don't see that that equality holds. The ideal $(x^n -2)$ consists of all polynomials in $x^n$, not all polynomials in $x$. So specifically, $\mathbb{C}[x]/(x^n-2)$ is an $n$-dimensional vector space over $\mathbb{C}$. – Craig Sep 8 '11 at 16:55
@rattle: $x^n - 2$ is not irreducible, so $(x^n - 2)$ is not maximal. – Qiaochu Yuan Sep 8 '11 at 17:39

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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