2
$\begingroup$

Let X and Y are irreducible spaces, do we have

$$\dim (X\times Y) = \dim X + \dim Y ?$$

We know that it is true for irreducible varieties, but here $X$ and $Y$ are only topological spaces.

  • 0
    Are you assuming that the spaces are Noetherian? If not, what notion of dimension do you have in mind?2017-01-20
  • 0
    The dimension is defined by the length of irreducible closed subset, the definition is in Hartshorne's book and it do not need the space is Noetherian. If necessary, we can add the Noetherian condition.2017-01-20
  • 0
    OK. Just to explain my earlier comment: This notion of the dimension of a topological space is one of many notions (see also small inductive dimension, large inductive dimension, and Lebesgue covering dimension). It is particularly well-suited to Noetherian spaces, and typically not that useful outside that context (for example, every Hausdorff space has dimension $0$, since the only irreducible sets are points). So when talking about the dimension of a bare topological space, with no other context, you should always specify what you mean by dimension.2017-01-20
  • 0
    Oh, maybe I should clear the definition of dimension, I think "irreducible" refers this dimension is from algebriac geometry before.2017-01-21
  • 1
    @Strongart 啃老智障,应该用elaborate而非clear!2017-05-01

2 Answers 2

3

In this answer, I'll use the definition of dimension given in the comments: $\dim(X)$ is the supremum of integers $n$ such that there exists a chain of length $n+1$ of irreducible closed subsets of $X$: $A_0\subsetneq A_1\subsetneq\dots \subsetneq A_n$. If there are arbitrarily long such chains, then $\dim(X) = \infty$.

With this definition, $\dim(X\times Y) = \dim(X) + \dim(Y)$ is true for any nonempty topological spaces $X$ and $Y$. To prove this, we'll first characterize the irreducible closed sets in the product topology.

Lemma: A set $Z\subseteq X\times Y$ is closed and irreducible in the product topology if and only if it has the form $A\times B$, where $A$ and $B$ are closed irreducible subsets of $X\times Y$.

Proof: In one direction, suppose $A\subseteq X$ and $B\subseteq Y$ are irreducible closed sets. Then $A\times B$ is closed - it is the compliment of the open set $(A^c\times Y)\cup (X\times B^c)$. We want to show it's irreducible. I think this is easiest to see if we use the formulation that a closed set $C$ is irreducible if and only for any open sets $U_1$ and $U_2$, if $U_1\cap C$ and $U_2\cap C$ are nonempty, then $U_1\cap U_2\cap C$ is nonempty. In fact, it suffices to check this condition for basic open sets, which in the product topology can be written $U_1\times V_1$ and $U_2\times V_2$. If $(U_i\times V_i)\cap (A\times B)$ is nonempty for $i = 1$ and $2$, then $U_i\cap A$ is nonempty for $i = 1$ and $2$, so there is some $x\in U_1\cap U_2\cap A$, since $A$ is irreducible. Similarly, there is some $y\in V_1\cap V_2\cap A$. But then $(x,y)\in (U_1\times V_1) \cap (U_2\times V_2)\cap (A\times B)$.

In the other direction, suppose $Z\subseteq X\times Y$ is closed and irreducible. Let $A = \overline{\pi_X(Z)}$ (the closure of the projection of $Z$ to $X$), and let $B = \overline{\pi_Y(Z)}$. Passing to the subspaces $A$ and $B$, we have that $Z\subseteq A\times B$ is closed and irreducible. Suppose for contradiction that it is a proper subset. Then there is some basic open set $U\times V\subseteq Z^c = (A\times B)\setminus Z$, with $U\subseteq A$ and $V\subseteq B$ open. And since $Z\subseteq (U^c\times B)\cup (A\times V^c)$ is irreducible, $Z$ is contained in one of these sets. If $Z\subseteq U^c\times B$, then $\pi_X(Z)\subseteq U^c$, contradicting the fact that $A = \overline{\pi_X(Z)}$, and similarly if $Z\subseteq A\times V^c$. $\square$

On to dimensions. Suppose we have a chain $A_0\subsetneq A_1\subsetneq\dots \subsetneq A_n$ witnessing $\dim(X)\geq n$ and a chain $B_0\subsetneq B_1\subsetneq\dots \subsetneq B_m$ witnessing $\dim(Y)\geq m$. Then the chain $$A_0\times B_0\subsetneq A_1\times B_0\subsetneq \dots\subsetneq A_n\times B_0\subsetneq A_n\times B_1\subsetneq \dots \subsetneq A_n\times B_m$$ witnesses $\dim(X\times Y)\geq n+m$. Indeed, all these sets are irreducible and closed, by the lemma. This shows $\dim(X\times Y)\geq \dim(X)+\dim(Y)$.

We'll prove the other inequality, $\dim(X\times Y)\leq \dim(X)+\dim(Y)$, by induction. The base case, $\dim(X\times Y) = 0$, is clear: the closure of a point is always irreducible, so any nonempty set has dimension at least $0$. Now suppose we have a chain $Z_0\subsetneq \dots \subsetneq Z_{n+1}$ witnessing $\dim(X\times Y) = n+1$. Then the subspace $Z_n$ has dimension $n$, and $Z_n = A\times B$ for $A\subseteq X$ and $B\subseteq Y$ closed and irreducible, by the lemma. By induction, $\dim(A) + \dim(B) \geq n$, witnessed by chains $A_0\subsetneq \dots\subsetneq A_i = A$ and $B_0\subsetneq \dots \subsetneq B_j = B$, with $i+j\geq n$. But we can also write $Z_{n+1} = A'\times B'$ for $A'$ and $B'$ closed irreducible, and since $Z_n\subsetneq Z_{n+1}$, we must have $A\subsetneq A'$ or $B\subseteq B'$. Hence at least one of the two chains (in $X$ or in $Y$) can be extended by at least one, and $\dim(X) + \dim(Y) \geq i+j + 1 \geq n+1$.

A final note: The point of this argument is that we don't get any interesting new irreducible closed sets in the product topology. Since the question mentions varieties, it's important to remember that if $X$ and $Y$ are algebraic varieties with the Zariski topology, and $X\times Y$ is the product of varieties, then the Zariski topology on $X\times Y$ is not the product topology. In fact, we get lots of new irreducible closed sets, e.g. all the plane curves in $\mathbb{A}^2 = \mathbb{A}^1 \times \mathbb{A}^1$ which aren't just lines parallel to the axes.

  • 0
    Thank you for the nice answer. In the final note, does it should be no interesting new irreducible closed sets in the Zariski topology? For the algebraic varieties, at least when k is algebraic closed, we have the irreducible closed subsets by minimal step, using a suitable hypersurface intersection, choosing a irreducible component, the dimension only reduces one for each step. Do this for X and Y respective and put them together, we have a chain for product topology, and there is no longer chain for Zariski topology, so there is the same dimension for product topology and Zariski topology.2017-01-22
  • 0
    I'm not sure I understand your comment, but it looks like you're outlining the proof that dimension is also additive for the product of varieties. In the product of topological spaces, there are no nontrivial new irreducible closed sets. In the product of varieties, there are *lots* of new nontrivial irreducible closed sets, but they don't screw up the dimension - the poset of irreducible closed sets is wider but not taller.2017-01-22
  • 0
    In the varieties case, we can count some numbers. A hypersurface intersection means add an irreducible equation. Given an algebraic variety with dimension n, we can add n irreducible equations step by step, until it becomes a point. Do you think it is the minimal possibility for each step? Now I do this for the product topology and treat it as for the Zariski topology, the number will not change.2017-01-23
0

Sorry, I don't have enough reputation to make this a comment. Are you sure that $\mathop{dim}(X \times Y)=\mathop{dim} X+\mathop{dim}Y$ even for irreducible varieties? This is certainly true if we take the product as varieties, but the topology on that product variety will typically not agree with the product topology.

  • 0
    Different topology but the same dimension, please see Alex Kruckman's answer and comment.2017-01-22