4

If $X\subset \mathbb{A}^N$ and $Y\subset\mathbb{A}^M$ are affine varieties with $X=Z(f_1,\dots,f_n)$ and $Y=Z(g_1,\dots,g_m)$ then $X\times Y\subset\mathbb{A}^{N+M}$ is an affine variety with $X\times Y=Z(f_1,\dots,f_n,g_1,\dots,g_m)$. But if we take $f_1,\dots,f_n$ and $g_1,\dots,g_m$ to generate the ideal of $X$ and $Y$ respectively then must $f_1,\dots,f_n,g_1,\dots,g_n$ generate the ideal of $X\times Y$?

This is equivalent to asking if given $f_1,\dots,f_n\in\mathbb{k}[x_1,\dots,x_n]$ and $g_1,\dots,g_n\in\mathbb{k}[y_1,\dots,y_m]$ generating radical ideals whether $f_1,\dots,f_n,g_1,\dots,g_n\in\mathbb{k}[x_1,\dots,x_n,y_1,\dots,y_m]$ generate a radical ideal.

This is true, for example, if the ideals are monomial ideals. But in general I don't really have good intuition for this and the main reason I am asking is because it would make working with product varieties a lot easier.

Seth
  • 9,563

2 Answers2

2

If $k$ is algebraically closed, then for two $k$-algebras $A$ and $B$ which are also integral domains, we will have $A\otimes_kB$ is an integral domain. (Reference.)

Let $X=\{x_1,\ldots,x_n\},Y=\{y_1,\ldots,y_m\}$. Now suppose $I,J$ are radical ideals in $k[X]$ and $k[Y]$ respectively, and suppose $k$ is algebraically closed. We will prove that $(I,J)$ is a radical ideal in $k[X,Y]$.

There are finitely many prime ideals $\mathfrak{p}_i\subseteq k[X]$, $\mathfrak{q}_j\subseteq k[Y]$ respectively such that $I=\cap_i \mathfrak{p}_i$ and $J=\cap_j \mathfrak{q}_j$. Thus we have injections $k[X]/I\to \prod_ik[X]/\mathfrak{p}_{i}$ and $k[Y]/J\to \prod_jk[Y]/\mathfrak{q}_j$. Therefore, we have an injection $k[X]/I\otimes_k k[Y]/J\to (\prod_ik[X]/\mathfrak{p}_i)\otimes (\prod_jk[Y]/\mathfrak{q}_j)\cong \prod_{i,j}k[X,Y]/(\mathfrak{p}_i,\mathfrak{q}_j)$. Since $\prod_{i,j}k[X,Y]/(\mathfrak{p}_i,\mathfrak{q}_j)$ is reduced, $k[X,Y]/(I,J)=k[X]/I\otimes k[Y]/J$ is also reduced.

Edit For the injectivity of $k[X]/I\otimes_k k[Y]/J\to (\prod_ik[X]/\mathfrak{p}_i)\otimes (\prod_jk[Y]/\mathfrak{q}_j)$:

(a) $k[X]/I\to \prod_ik[X]/\mathfrak{p}_i$ is injective, tensoring $k[Y]/J$ gives that $k[X]/I\otimes_kk[Y]/J\to (\prod_ik[X]/\mathfrak{p}_i)\otimes_kk[Y]/J$ is injective.

(b)$k[Y]/J\to \prod_jk[Y]/\mathfrak{q}_j$ is injective, tensoring $\prod_ik[X]/\mathfrak{p}_i$ gives $(\prod_ik[X]/\mathfrak{p}_i)\otimes_kk[Y]/J\to (\prod_ik[X]/\mathfrak{p}_i)\otimes (\prod_jk[Y]/\mathfrak{q}_j)$ is injective.

Then the composition of two injections is injective.

user119882
  • 1,540
  • Thank you very much for your answer. Apparently my algebra background is a lot weaker than I thought it was. I am unsure about a few of the steps. Is it true that tensoring two injective maps (over a field) preserves injectivity? By reduced do you mean has no nilpotents? (I'm not sure why) Also I don't understand k[X,Y]/(I,J)=k[X]/I⊗k[Y]/J or (∏ik[X]/pi)⊗(∏jk[Y]/qj)≅∏i,jk[X,Y]/(pi,qj) (these look like two different cases of some general property I probably don't know about. (sorry for asking so many questions!!) – Seth Mar 29 '14 at 17:27
  • Ok I think tensoring over field preserves injectives because vector spaces are always flat modules. – Seth Mar 29 '14 at 17:30
  • 1
    You need to show $k[X]\otimes_k k[Y]=k[X,Y]$, and use the fact $A/I\otimes_AM\cong M/IM$ for any ring $A$, an $A$-module $M$, an ideal $I$ of $A$, then you can conclude that $k[X]/I\otimes_kk[Y]/J=(k[X]/I\otimes_{k[X]}(k[X]\otimes_kk[Y]))\otimes_{k[Y]}k[Y]/J=k[X,Y]/Ik[X,Y]\otimes_{k[Y]}k[Y]/J=k[X,Y]/(I,J)$. Clearly, those isomorphisms are isomorphisms as modules, but they are also $k$-algebra homomorphisms, so, they are isomorphisms as rings.

    We have an isomorphism $(A\times B)\otimes C\cong (A\otimes C)\times (B\otimes C)$ as modules, then check the isomorphism is indeed a ring map.

    – user119882 Mar 29 '14 at 18:25
  • Thank you very much, I hate to bother you again, but why are is ∏i,jk[X,Y]/(pi,qj) reduced? – Seth Mar 29 '14 at 18:36
  • Yes, you are right. Tensoring over a field preserves injectives, vector spaces are free. And, a ring is said reduced if it has no nilpotents. @Seth – user119882 Mar 29 '14 at 18:39
  • Since all $k[X,Y]/(\mathfrak{p}_i,\mathfrak{q}_j)$ are domains by the first sentense of the answer. You are welcome. – user119882 Mar 29 '14 at 18:41
  • k[X,Y] is an integral domain but why is ∏i,jk[X,Y]/(pi,qj) reduced? – Seth Mar 29 '14 at 19:03
  • @user26857 I believe it does follow from flatness since both sides are flat modules and if $N$ is flat then basically $f\otimes 1$ is injective if $f$ is. (so do this twice on left argument and right argument) (ok this notation makes no sense but I think you see what I'm saying, I'm referencing a theorem from atiyah macdonald) – Seth Mar 29 '14 at 19:04
  • @user26857 yes, thank you. That argument would be simpler (although I need to find a proof of that result somewhere then) – Seth Mar 29 '14 at 19:05
  • Although proving the tensor product of two domains is a domain is something we actually proved in class (using algebraic geometry) – Seth Mar 29 '14 at 19:05
  • @user26857 Thanks, I will go look up the theorem. Somehow I feel like what I am trying to prove is almost equivalent to the tensor product of two reduced algebras is reduced, for a certain subclass of algebras. – Seth Mar 29 '14 at 19:10
  • @user26857 I read the proof, it was very nice. I wish we were using these notes instead of Shafarevich. Milne seems much more to my taste. – Seth Mar 29 '14 at 19:23
  • Ok I thought about this some more and over an algebraically closed field tensor product of two reduced algebras is a reduced algebra is basically the same as the product of varieties is a variety. Tensor product of two domains is a domain is basically the same as product of two irreducible varieties is an irreducible variety. I'm still confused about the proof of the result I'm looking for. I will continue to think about it. – Seth Mar 29 '14 at 19:35
  • @user26857, I read the proof of Milne's, it is great, thanks. – user119882 Mar 29 '14 at 19:40
  • @user119882 I just realized I was misinterpretting your notation the whole time and now I almost understand what you are saying... (lol) – Seth Mar 29 '14 at 19:43
  • @user26857 If you'd like to that would be great (but totally up to you). I just posted a basic proof which solves the problem for me. Proposition 2.19 in Atiyah Macdonald says: If an $A$-module $N$ is flat then given $f:M'\to M$ injective which $M$, $M'$ finitely generated then $f\otimes 1:M'\otimes N \to M \otimes N$ is injective. The symmetric proposition must also be true so we compose the two injective maps. In your example I think the ring hom an embedding to a subring without zero divisors. – Seth Mar 29 '14 at 20:12
2

A simpler answer:

Since $I=(f_1,\dots, f_n)$ and $J=(g_1,\dots,g_m)$ are radical ideals, then $k[X]/I$ and $k[Y]/J$ are reduced $k$-algebras. If we assume that $k$ is a perfect field (in particular, algebraically closed) we get that $k[X]/I\otimes_kk[Y]/J=k[X,Y]/(I,J)$ is also reduced (see the last part of this answer for perfect fields or Milne's book Algebraic Geometry, Proposition 4.15(a) for algebraically closed fields), and thus in this case the answer to the question is positive.

user26857
  • 53,190
  • Thanks for the answer. I just realized that I still don't understand k[X]/I⊗kk[Y]/J=k[X,Y]/(I,J). I just went to write it down on paper and realized I can't justify all the steps. – Seth Mar 29 '14 at 21:04
  • 1
    @Seth Try to take a look here. – user26857 Mar 29 '14 at 21:16