Categorical similarities between Galois theory and Hilbert's Nullstellensatz
问题内容
In the following, I am going to compare two correspondences: $\textbf{correspondence between intermediate fields of $L/K$ and subgroups of ${\rm Aut}_K(L)$}$ vs $\textbf{correspondence between ideals of $R=k[x_1,\cdots,x_n]$ and algebraic subsets of $\mathbb{A}^n_k$}.$
While doing this, I will show you some similarities, which, to me, are more than a coincidence. Then, I will ask you how we are able to explain such similarities from a broader perspective--maybe a categorical explanation.
Consider the following four maps: $$\mathcal{Z}:R \to \mathbb{A}^n_k$$ sending subset $T$ of the polynomial ring over algebracially closed field to their common zeros $\mathcal{Z}(T)$, $$\mathcal{I}:\mathbb{A}^n_k \to R$$ sending subset $Y$ of the affine space with Zariski topology to the radical ideal whose zero set is $\bar{Y}$, $$\mathcal{G}:L \to {\rm Aut}_K(L)$$ sending intermediate field $E$ to $\mathcal{G}(E)={\rm Aut}_E(L)$, and $$\mathcal{F}:{\rm Aut}_K(L) \to L$$ sending subgroup $H$ to the fixed field $\mathcal{F}(H)=L^H$. Note also that we extend $\mathcal{F}$ to send a subset $S$ to the fixed field $\mathcal{F}(\langle S \rangle)$ of the smallest subgroup $\langle S\rangle$ containing $S$. Observe the following similarities:
Similarity 1:
We have \begin{align}\mathcal{ZIZ}&=\mathcal{Z}\\ \mathcal{GFG}&=\mathcal{G}.\end{align} For any ideal $a \subset R$, $\mathcal{ZIZ}(a)=\mathcal{Z}(a)$. Similarly, for any intermediate field $E$, $\mathcal{GFG}(E)=\mathcal{G}(E)={\rm Aut}_E(L)$.
Similarity 2:
We have \begin{align}\mathcal{IZI}&=\mathcal{I}\\ \mathcal{FGF}&=\mathcal{F}.\end{align} For any subset $Y \subset \mathbb{A}^n_k$, $\mathcal{IZI}(Y)=\mathcal{I}(Y)=\sqrt{\mathcal{I}(Y)}$. Similarly, for any subset $S \subset {\rm Aut}_K(L)$, $\mathcal{FGF}(S)=\mathcal{F}(S)=L^{\langle S\rangle}$.
Similarity 3:
For any subsets $Y \subset \mathbb{A}^n_k$ and $S \subset {\rm Aut}_K(L)$, we have \begin{align}\mathcal{ZI}(Y)&\supset Y\\ \mathcal{GF}(S)&\supset S.\end{align}
Similarity 4:
For any ideal $a \subset R$ and intermediate field $L/E/K$, we have \begin{align}\mathcal{IZ}(a)&\supset a\\ \mathcal{FG}(E)&\supset E.\end{align}
Similarity 5:
If we resterict ourselves to closed subsets $Y \subset \mathbb{A}^n_k$ and subgroups $ S \subset {\rm Aut}_K(L)$, then we have \begin{align}\mathcal{ZI}&=id\\ \mathcal{GF}&=id.\end{align}
Similarity 6:
If we resterict ourselves to radical ideals $a \subset R$ and Galois extensions $L/E$, then we have \begin{align}a=\sqrt{a} &\Longrightarrow &\mathcal{IZ}&=id\\ L^{{\rm Aut}_E(L)}=E &\Longrightarrow &\mathcal{GF}&=id.\end{align}
I know that Similarities 5 and 6 are just special cases of the former similarities, but still I wanted to distinguish them. It is like the subsets of $\mathbb{A}^n_k$ behave similarly to the subsets of ${\rm Aut}_K(L)$, and the algebraic subsets of $\mathbb{A}^n_k$ behave similarly to the subgroups of ${\rm Aut}_K(L)$. Furthermore, ideals behave like extensions $L/E$ and radical ideals behave like Galois extensions $L/E$. How do you explain all these similarities? To me, it's like one part of math copy-pasted with adjustments to another part. Is there a categorical explanation for this? Something like $\mathcal{Z}$ and $\mathcal{G}$ are functors with very similar features, and also $\mathcal{I}$ and $\mathcal{F}$ their inverse functors again with similar features?
回答 (1)
You are close to rediscovering the concept of Galois connections. This concept formalizes the "order-theoretic essence" of situations like those two that you describe. In the blogpost linked in the comment and the wikipage linked in this answer you can find more examples. Galois connections are ubiquitous in mathematics, probably you can come up with more examples.
Category theory further generalizes the concept of a Galois connection to that of adjoint functors. A poset can be thought of as a small thin category and if we apply the concept of adjunctions to this case, we recover Galois connections. Many properties of Galois connections have some analogue for adjunctions.
For example, any adjunction $F \vdash G$ induces an equivalence of categories between the "fixed points", i.e. between the objects $X$ such that the natural morphisms $FG (X) \to X$ is an isomorphism and between the objects $Y$ such that $Y \to GF(Y)$ is an isomorphism.
The equalities $GFG=G$ and $FGF=F$ generally do not hold,instead we have more complicated triangle identities.