2

My current understanding: this is always true

Let $A$ be a ring. Let $S$ denote a subset of the ring. Let $V(S)$ be the set of prime ideals of $A$ containing $S$. Define closed sets in $X= \operatorname{Spec} (A)$ to be those of the form $V(S)$.

As I understand, to any radical ideal $I$, we can uniquely associated the closed set $V(I)$. Conversely, to any closed set $V$, we can associate the radical ideal given by the intersection of all elements of $V$.

I can provide proofs I have written to these if this helps, I haven't added them now to avoid drowning my question.

My worry

I haven't been able to find anything on closed sets - radical ideal correspondence, except when we are specifically with algebraic sets in $\mathbb{A}^n_K$ and radical ideals in $K[X_1, ..., X_n]$, such as on this page linked in this question. I understand this is a fundamental result in algebraic geometry (which I am only just beginning to study), but I fail to see why the result may not be more general.

My question

Is my current understanding correct, or have I made a mistake such as forgetting a fringe case ?

RobPratt
  • 50,938
Mr Lolo
  • 463

2 Answers2

6

You're correct, there is a bijective correspondence between closed subsets of $\operatorname{Spec} A$ and radical ideals $I\subsetneq A$ and this always holds.

This correspondence gets discussed a lot in "classical" algebraic geometry (the typical/original formulation of Hilbert's Nullstellensatz, for instance) because "classical" algebraic geometry can't distinguish between all the different ideals which have the same set of zeroes - $V(x^2)=V(x)$ as affine varieties in $\Bbb A^n$ to any algebraic geometer in the 1800s, for instance. Part of the importance of upgrading to schemes is that now algebraic geometry can distinguish $V(x^2)$ and $V(x)$ because they have different subscheme structures, and that's usually information we keep track of. So it's slightly less important to talk about the underlying closed set.

That doesn't mean we completely forget about that, though - for any closed subset $Y$ in any scheme $X$, there is a unique way to turn this in to a closed subscheme called the reduced induced subscheme structure. Affine-locally this corresponds to exactly what you're talking about here - given a closed subset which can be written as the set of points underlying $V(I)\subset \operatorname{Spec} A$, consider it as the subscheme $V(\sqrt{I})$.

KReiser
  • 74,746
0

A detour into conceptual development may help.

Before having the language of affine scheme/space $\text{Spec }A$ and related notions of varieties, the classical algebraic geometry treats algebraic varieties simply as subsets of $k^n$ of the form $$V(J) := \{\mathbf{x} \in k^n : f(\mathbf{x}) = 0 \text{ for all } f \in J \},$$ where $J$ is an ideal of ring $k[x_1,\ldots,x_n]$. The closed set - radical ideal correspondence in the classical setting is concluded by Nullstellensatz, which states that $I(V(J)) = \sqrt{J}$, with a non-trivial proof of course.

This correspondence became a source of inspiration for redefining and expanding the definition of (affine & algebraic) variety in a way more algebraically (and less dependent on the readily built-in vector space structure of $k^n$), which ends up in the formulation of $\text{Spec } A$ by prime ideal and new Zariski-topology with new notion of closed set that you posed. This definition does not immediately seem to coincide with the classical setting, but again that is because the technical gap is basically occupied by Nullstellensatz. Now with the new definition, the "Nullstellensatz-like" property of $I(V(J)) = \sqrt{J}$, as you've said in your post, is an immediate check, hence we can almost say such a correspondence is "by definition". There is more to this new setting of course, for example as suggested by KReiser's answer of the ability to differentiate the subscheme structures of different ideals.

Beyond the original post, I'd share another interesting discussion in the nuances behind the concept of affine space.

eiKeViN
  • 26