5

Let $k$ be an algebraically closed field, and $A$ be a finitely generated $k$-algebra with no nilpotents.

Show that $\operatorname{Spec}A$ is a finite set if and only if $A$ is a finite dimensional vector space over $k$.

egreg
  • 238,574

3 Answers3

7

Suppose $A$ is a finitely generated algebra (not necessarily reduced) over a field $k$ (not necessarily algebraically closed).
Then Noether's normalization theorem says that there exist $n\geq 0$ elements $y_1,\cdots, y_n\in A$, algebraically independent over $k$ , such that $A$ is module-finite over its sub-algebra $k[y_1,\cdots,y_n]$.
Since $\text {Spec} (A)\to \text {Spec} (k[y_1,\cdots, y_n])=\mathbb A^n_k$ is surjective by finiteness and since $\mathbb A^n_k$ is an infinite set for $n\gt 0$, it follows that $\text {Spec} (A)$ is a finite set if and only if $n=0$ i.e. if and only if $A$ is finite dimensional over $k$.

4

Let me try an alternative proof.

If Spec $A$ is finite, then there are only finitely many maximal ideals, say $m_1,\dots, m_n$. On the other hand, since $A$ is a f.g. $k$-algebra, it follows from Nullstellensatz that nil $A:=\sqrt{(0)}=\bigcap$ Specm $A$, where Specm $A$ denotes the set of maximal ideals of $A$. Thus, there exists $l\in \mathbb{N}$ such that $(m_1\dots m_n)^l=(0)$. Now it is easy to conclude that $A$ has finite lenght as $k$-module, that is, $\dim_k A<\infty$.

Conversely, $\dim_k A<\infty$ implies that $A$ is artinian, hence $\dim A=0$. But noetherian rings with dimension $0$ have discrete spectrum (because they have only finitely many minimal prime ideals, which must also be maximal by dimension). Finally, since Spec $A$ is quasi-compact, if Spec ${A}$ is discrete, then it is finite.

3

As $A$ is a finitely generated $k$-algebra, we have a surjection $k[x_1,\ldots x_r] \to A$ and thus $A \cong k[x_1,\ldots ,x_r]/I$. Assume that $Spec A$ is finite, then there are only finitely many maximal ideals containing $I$ in $k[x_1, \ldots ,x_r]$, say $m_1, \ldots m_n$. As these are co-maximal, we have $\cap m_i = \prod m_i$ and thus this product contains $I$, hence we have a map: $k[x_1, \ldots, x_r]/I \to k[x_1, \ldots, x_r]/\prod m_i$. Using The Chinese remainder theorem ,the right hand side is equal to $\oplus k[x_1, \ldots, x_r]/m_i$, and using the nullstellensatz + that $k$ is algebraically closed, we obtain that this is again equal to $\oplus_{i=1}^n k$ which is a finite dimensional vector space over $k$. The map above is obviously surjective, we now check injectivity: If $f$ is mapped to $0$ then, $f$ is contained in $\prod m_i = \cap m_i$, but as $\cap m_i = \cap P = \sqrt I$ in $k[x_1,\ldots x_r]$ (here the intersections are taken over maximal ideals containing $I$ and prime ideals containing $I$ respectively), it follows that the image of $f$ in $k[x_1,\ldots, x_r]/I$ is contained in the nilradical , but this is isomorphic to $A$ which was assumed not to have any nilpotents, thus the image of $f = 0$ and the map is an injection aswell. Conversely: Assume now that $A = k[x_1, \ldots, x_r]/I$ is a finite dimensional vector space over $k$ and assume for the sake of contradiction that $spec A$ is infinite, then given any $n$ we can find $n$ different prime ideals each containing $I$ , say $P_1, \ldots P_n$ and we have a surjection $k[x_1, \ldots x_r]/I \to k[x_1, \ldots x_r]/\prod P_i = \oplus k[x_1,\ldots x_r]/P_i$. The right hand side has dimension over $k$ at least equal to $n$, and thus $A$ has dimension greater or equal to $n$ for arbitrary $n$. This is a contradiction, thus $Spec A$ must be finite. Added: The reason why we can use the CRT on $k[x_1,\ldots x_r]/\prod P_i$, is because as $A$ is a finite dimensional vector space over $k$ the descending chain condition holds for sub-k-modules (and hence also for sub-A-modules i.e. the ideals of $A$), thus $A$ is Artin, and in an Artin ring we have that every prime ideal is maximal.

  • I might have been a bit sloppy in explaining maps and notation, and in the converse I don't explain as much as many of the arguments are the same as above. Feel free to ask if there is something in my proof that is unclear, and everyone is welcome to edit this. – Oliver E. Anderson Mar 05 '14 at 16:14
  • I also think you don't need $k$ to be algebraically closed as $k[x_1,\ldots,x_n]/m_i$ will be a finite algebraic field extension which always is finite dimensional over $k$. – Oliver E. Anderson Mar 05 '14 at 16:50