3
$\begingroup$

Let $C$ be an algebraically closed field and consider a variety $X$ over $C$. In the language of schemes $X$ is a separated, integral scheme over $C$ with a morphism of finite type $f:X\longrightarrow \operatorname{Spec} C$. If $\sigma\in\operatorname{Aut} C$ then we can define the variety $X^\sigma$ as the scheme $X$ but with a morphism of finite type $\operatorname{Spec}(\sigma)\circ f: X\longrightarrow \operatorname{Spec} C$. So $X$ and $X^\sigma$ are the same as schemes but not as varieties.

Now I'm trying to investigate the relationship between $X$ and $X^\sigma$ aside from the scheme theory, but in the language of classical algebraic geometry. In literature I've found the following theorem with a very short proof that I don't understand.

Theorem: Let $V(f_1,\ldots,f_m)\subset\mathbb P^n_C$ a classical variety where $f_1,\ldots,f_m$ are homogeneous polynomial of $C[x_0,\ldots,x_n]$. If $\sigma\in\operatorname{Aut(C)}$, and with an abuse of notation we also indicate with $\sigma$ the induced authomorphism on $C[x_0,\ldots,x_n]$, then the closed subset $V(\sigma^{-1}(f_1),\ldots,\sigma^{-1}(f_m))$ "corresponds" to $X^\sigma$ (defined above) in the category of classical varieties. To be more precise if we apply the usual functor to the category of schemes at $V(f_1,\ldots,f_m)$ we obtain a variety $X$ and if we apply the same functor at $V(\sigma^{-1}(f_1),\ldots,\sigma^{-1}(f_m))$ we obtain $X^\sigma$.

Proof: The two graded algebras $\frac{C[x_0,\ldots,x_n]}{(f_1,\ldots,f_m)}$ and $\frac{C[x_0,\ldots,x_n]}{(\sigma^{-1}(f_1),\ldots,\sigma^{-1}(f_m))}$ are clearly isomorphic thanks to an isomorphism $\widetilde\sigma$ induced by $\sigma$. Using the morphism $\operatorname{Proj}(\widetilde\sigma)$ follows easily the thesis.


Practically the theorem explains what is $X^\sigma$ for concrete varieties such as closed subsets of $\mathbb P^n_C$, but I don't understand the last sentence of the proof; in particular applying the functor $\operatorname{Proj}$, I see only the isomorphism between $\operatorname{Proj}\left(\frac{C[x_0,\ldots,x_n]}{(f_1,\ldots,f_m)}\right)$ and $\operatorname{Proj}\left(\frac{C[x_0,\ldots,x_n]}{(\sigma^{-1}(f_1),\ldots,\sigma^{-1}(f_m))}\right)$. Maybe it is a stupid question, since the proof should follow easily, but please help me to understand it.

Thanks in advance.


References: Essentially two articles: "B.Köck - Belyi's theorem revisited" and "H.Hammer, F.Herrlich - a remark on the moduli field of a curve "

$\endgroup$
  • 1
    $\begingroup$ Watch out: Proj is not a functor! $\endgroup$ – Bruno Joyal Dec 29 '13 at 21:31
2
$\begingroup$

You can view it in this way: think of $\sigma$ as $\sigma:C'\to C$ and forget that $C'=C$, then $X^\sigma\rightarrow\operatorname{Spec}(C)\xrightarrow{\sigma}\operatorname{Spec}(C')$ is a $C'$ scheme. Then the following diagram is cartesian by easily checking the universal property

\begin{array}[c]{ccc} X^\sigma&=&X\\ \downarrow&\square &\downarrow\\ \operatorname{Spec}(C')&\xrightarrow{\operatorname{Spec}(\sigma^{-1})}& \operatorname{Spec}(C) \end{array}

It's well known that, if $X=\operatorname{Proj} A$ as a $C$ scheme, then $X^\sigma=X\times_{\operatorname{Spec}(C)}\operatorname{Spec}(C')\simeq\operatorname{Proj}(A\otimes_CC')$ canonically as a $C'$ scheme. So, if $$A=C[x_0,\ldots,x_n]/(f_1,\ldots,f_m)$$ then $$A\otimes_CC'=C'[x_0,\ldots,x_n]/(\sigma^{-1}(f_1),\ldots,\sigma^{-1}(f_m))$$ Oh, but now I remember: $C'=C$!

| cite | improve this answer | |
$\endgroup$
2
$\begingroup$

This is not a complete answer, but I will describe what happens in the affine case. Suppose we have polynomials $f_1,\ldots,f_m\in C[y_1,\ldots,y_n]$ and an automorphism $\sigma: C\rightarrow C$. Then we have a commutative diagram $$\begin{array}[c]{ccc} C&\rightarrow&\frac{C[y_1,\ldots,y_n]}{(f_1,\ldots,f_m)}\\ \uparrow\scriptstyle{\sigma}& &\uparrow\scriptstyle{\sigma}\\ C&\rightarrow& \frac{C[y_1,\ldots,y_n]}{(\sigma^{-1}f_1,\ldots,\sigma^{-1}f_m)} \end{array} $$ which induces a conmutative diagram $$\begin{array}[c]{ccc} Spec(C)&\leftarrow&Spec\left(\frac{C[y_1,\ldots,y_n]}{(f_1,\ldots,f_m)}\right)\\ \downarrow\scriptstyle{\sigma}& &\downarrow\scriptstyle{\sigma}\\ Spec(C)&\leftarrow& Spec\left(\frac{C[y_1,\ldots,y_n]}{(\sigma^{-1}f_1,\ldots,\sigma^{-1}f_m)}\right). \end{array} $$ Then $\sigma:Spec\left(\frac{C[y_1,\ldots,y_n]}{(f_1,\ldots,f_m)}\right)\rightarrow Spec\left(\frac{C[y_1,\ldots,y_n]}{(\sigma^{-1}f_1,\ldots,\sigma^{-1}f_m)}\right)$ is an isomorphism between the $C$-scheme $Spec\left(\frac{C[y_1,\ldots,y_n]}{(f_1,\ldots,f_m)}\right)\rightarrow Spec(C)\xrightarrow{\sigma}Spec(C)$ and the $C$-scheme $ Spec\left(\frac{C[y_1,\ldots,y_n]}{(\sigma^{-1}f_1,\ldots,\sigma^{-1}f_m)}\right)\rightarrow Spec(C)$, that is, between $V(f_1,\ldots,f_n)^\sigma$ and $V(\sigma^{-1}f_1,\ldots,\sigma^{-1}f_m)$.

I think that the projective case should work the same as this or it could be deduced from this by glueging.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.