Core idea

User

Let's see, doesn't an argument of Grothendieck's give that a of is equivalent data to a ?

Assistant

Yes, with one important qualification: it is a in the , not usually in the .

Let K/FK/F be a and let

G=Gal(K/F)=AutF(K).G=\operatorname{Gal}(K/F)=\operatorname{Aut}_F(K).

Then the corresponding is

SpecKSpecF.\operatorname{Spec} K \longrightarrow \operatorname{Spec} F.

Here:

P=SpecK,M=SpecF,G=Gal(K/F).P=\operatorname{Spec}K,\qquad M=\operatorname{Spec}F,\qquad G=\operatorname{Gal}(K/F).

The GG on KK by FF-algebra , hence on SpecK\operatorname{Spec}K over SpecF\operatorname{Spec}F. This makes

SpecKSpecF\operatorname{Spec}K \to \operatorname{Spec}F

a , i.e. a , in the .

The key algebraic condition is the

KFKσGK,K\otimes_F K \cong \prod_{\sigma\in G} K,

given by

ab(aσ(b))σG.a\otimes b \mapsto \big(a\,\sigma(b)\big)_{\sigma\in G}.

Geometrically, this says

SpecK×SpecFSpecKSpecK×G.\operatorname{Spec}K \times_{\operatorname{Spec}F} \operatorname{Spec}K \cong \operatorname{Spec}K \times G.

That is the : over SpecK\operatorname{Spec}K, the bundle into one copy of the for each gGg\in G.

So the slogan is:

finite Galois extension K/Fconnected finite eˊtale G-torsor over SpecF.\boxed{\text{finite Galois extension } K/F \quad\leftrightsquigarrow\quad \text{connected finite étale } G\text{-torsor over } \operatorname{Spec}F.}

But more precisely:

K/F finite Galois with group GK/F \text{ finite Galois with group }G

determines a

SpecKSpecF\operatorname{Spec}K\to\operatorname{Spec}F

in the .

Conversely, a over SpecF\operatorname{Spec}F gives a finite Galois FF-algebra. If the is , that is a K/FK/F.

User

What is the ? Also how do I type the accented e in macOS?

Assistant

The is a on . It is not usually a topology on the X|X|. It is a rule for which into XX count as “.”

For a XX, define the XeˊtX_{\mathrm{\acute{e}t}} as follows.

:

UXU \to X

where UU is a and UXU \to X is an .

:

(VX)(UX)(V\to X)\to (U\to X)

are VUV\to U.

:

{UiU}iI\{U_i \to U\}_{i\in I}

such that each UiUU_i\to U is étale and

iim(UiU)=U.\bigcup_i \operatorname{im}(U_i\to U)=U.

So the slogan is:

eˊtale topology=topology where eˊtale maps are allowed as local charts/covers.\boxed{\text{étale topology}=\text{topology where étale maps are allowed as local charts/covers}.}

An is a that is:

locally of finite presentation+flat+unramified.\text{locally of finite presentation}+\text{flat}+\text{unramified}.

It is the algebraic-geometry analogue of a .

For , this becomes very concrete:

SpecKSpecF\operatorname{Spec}K\to \operatorname{Spec}F

is iff K/FK/F is finite separable. Therefore a K/FK/F gives an étale cover

SpecKSpecF.\operatorname{Spec}K\to\operatorname{Spec}F.

This is why the statement is made in the . In the ordinary , SpecF\operatorname{Spec}F has only one point, so there are no nontrivial that can trivialize SpecKSpecF\operatorname{Spec}K\to\operatorname{Spec}F. The adds algebraic covers like .

For typing é on macOS:

Option+e, then e\text{Option} + e,\ \text{then } e

So:

  1. Press Option + e
  2. Release
  3. Press e

For uppercase:

Option+e, then Shift+e\text{Option}+e,\ \text{then Shift}+e

giving É.

In LaTeX:

\'{e}tale

or directly:

étale

User

Sorry what is X and how is it related to K and F?

Assistant

In the general definition, XX is just an arbitrary .

For your example, take

X=SpecF.X=\operatorname{Spec}F.

Then the étale cover/ is

SpecKSpecF.\operatorname{Spec}K \longrightarrow \operatorname{Spec}F.

So in this example:

X=SpecF,P=SpecK,G=Gal(K/F).\boxed{ X=\operatorname{Spec}F,\qquad P=\operatorname{Spec}K,\qquad G=\operatorname{Gal}(K/F). }

The map

SpecKSpecF\operatorname{Spec}K \to \operatorname{Spec}F

comes from the

FKF \hookrightarrow K

by of Spec\operatorname{Spec}. In algebraic geometry, reverse direction:

FKF \hookrightarrow K

induces

SpecKSpecF.\operatorname{Spec}K \to \operatorname{Spec}F.

The reason I introduced a general XX was to define the on any . For a general XX, of XeˊtX_{\acute{e}t} are

UX.U\to X.

In the Galois case, the base is specifically

X=SpecF,X=\operatorname{Spec}F,

and one distinguished étale over it is

SpecKSpecF.\operatorname{Spec}K\to\operatorname{Spec}F.

If K/FK/F is finite separable, this map is . If K/FK/F is finite Galois with GG, it is a in the .

User

axiomatically define and give examples.

Assistant

A is a space locally modeled on .

Fix CRing\mathbf{CRing} = of .

1.

A is a pair

(X,OX)(X,\mathcal O_X)

where:

: XX is a .

: OX\mathcal O_X is a of on XX.

: For every xXx\in X, the

OX,x\mathcal O_{X,x}

is a , i.e. it has a unique .

The OX\mathcal O_X is the “” on XX, though its elements may be algebraic rather than ordinary .

2.

Given ACRingA\in \mathbf{CRing}, define

SpecA\operatorname{Spec} A

as follows.

Points:

SpecA={pA:p is a prime ideal}.|\operatorname{Spec} A| = \{\mathfrak p\subseteq A : \mathfrak p \text{ is a prime ideal}\}.

: For each IAI\subseteq A,

V(I)={pSpecA:Ip}.V(I)=\{\mathfrak p\in \operatorname{Spec}A : I\subseteq \mathfrak p\}.

These define the .

: For fAf\in A,

D(f)={pSpecA:fp}.D(f)=\{\mathfrak p\in \operatorname{Spec}A : f\notin \mathfrak p\}.

: Define

OSpecA(D(f))=Af,\mathcal O_{\operatorname{Spec}A}(D(f))=A_f,

where AfA_f is the of AA obtained by inverting ff. These assignments glue to a OSpecA\mathcal O_{\operatorname{Spec}A}.

: At pSpecA\mathfrak p\in \operatorname{Spec}A,

OSpecA,pAp.\mathcal O_{\operatorname{Spec}A,\mathfrak p}\cong A_{\mathfrak p}.

Since ApA_{\mathfrak p} is , SpecA\operatorname{Spec}A is a .

An is a to

(SpecA,OSpecA)(\operatorname{Spec}A,\mathcal O_{\operatorname{Spec}A})

for some ACRingA\in\mathbf{CRing}.

3.

A is a

(X,OX)(X,\mathcal O_X)

such that:

cover:

X=iIUiX=\bigcup_{i\in I} U_i

for UiXU_i\subseteq X, and for each ii, there exists a AiA_i such that

(Ui,OXUi)SpecAi(U_i,\mathcal O_X|_{U_i})\cong \operatorname{Spec}A_i

as .

So:

scheme=locally ringed space locally isomorphic to SpecA.\boxed{ \text{scheme}=\text{locally ringed space locally isomorphic to }\operatorname{Spec}A. }

Examples

1. One-point from a

If FF is a , then

SpecF={(0)}.\operatorname{Spec}F=\{(0)\}.

It has one point, and its has

OSpecF,(0)=F.\mathcal O_{\operatorname{Spec}F,(0)}=F.

This is the appearing in the example:

SpecKSpecF.\operatorname{Spec}K\to\operatorname{Spec}F.

2. The over a

For a kk,

Ak1:=Speck[x].\mathbb A^1_k := \operatorname{Spec} k[x].

Its points are of k[x]k[x]. For example, if kk is , the are

(xa)(x-a)

for aka\in k, plus the (0)(0).

3.

Akn:=Speck[x1,,xn].\mathbb A^n_k := \operatorname{Spec} k[x_1,\dots,x_n].

This is the version of nn-dimensional space over kk.

4. The

SpecZ.\operatorname{Spec}\mathbb Z.

Its points are

(0), (2), (3), (5), (7),(0),\ (2),\ (3),\ (5),\ (7),\dots

where (p)(p) ranges over generated by rational primes pp.

The point (0)(0) is the . The points (p)(p) are closed.

5.

over kk is

Pkn.\mathbb P^n_k.

It is not usually , but it is a because it has an

Pkn=i=0nUi,\mathbb P^n_k=\bigcup_{i=0}^n U_i,

with

UiAkn.U_i\cong \mathbb A^n_k.

So include both like SpecA\operatorname{Spec}A and glued from pieces, like Pkn\mathbb P^n_k.