Core idea
User
Let's see, doesn't an argument of Grothendieck's give that a Galois extension of fields is equivalent data to a principal Gal(K/F)-bundle?
Assistant
Yes, with one important qualification: it is a principal bundle in the étale topology, not usually in the ordinary topology.
Let K/F be a finite Galois extension and let
G=Gal(K/F)=AutF(K).
Then the corresponding principal G-bundle is
SpecK⟶SpecF.
Here:
P=SpecK,M=SpecF,G=Gal(K/F).
The group G acts on K by F-algebra automorphisms, hence contravariantly acts on SpecK over SpecF. This makes
SpecK→SpecF
a G-torsor, i.e. a principal G-bundle, in the étale site.
The key algebraic condition is the Galois identity
K⊗FK≅σ∈G∏K,
given by
a⊗b↦(aσ(b))σ∈G.
Geometrically, this says
SpecK×SpecFSpecK≅SpecK×G.
That is the torsor condition: over SpecK, the bundle trivializes into one copy of the fiber for each g∈G.
So the slogan is:
finite Galois extension K/F↭connected finite eˊtale G-torsor over SpecF.
But more precisely:
K/F finite Galois with group G
determines a principal G-bundle
SpecK→SpecF
in the étale topology.
Conversely, a connected finite étale G-torsor over SpecF gives a finite Galois F-algebra. If the total space is connected, that algebra is a field extension K/F.
User
What is the etale topology? Also how do I type the accented e in macOS?
Assistant
The étale topology is a Grothendieck topology on schemes. It is not usually a topology on the point-set ∣X∣. It is a rule for which morphisms into X count as “open covers.”
For a scheme X, define the small étale site Xeˊt as follows.
Objects:
where U is a scheme and U→X is an étale morphism.
Morphisms:
(V→X)→(U→X)
are X-morphisms V→U.
Covering families:
{Ui→U}i∈I
such that each Ui→U is étale and
i⋃im(Ui→U)=U.
So the slogan is:
eˊtale topology=topology where eˊtale maps are allowed as local charts/covers.
An étale morphism is a morphism of schemes that is:
locally of finite presentation+flat+unramified.
It is the algebraic-geometry analogue of a local diffeomorphism.
For fields, this becomes very concrete:
SpecK→SpecF
is finite étale iff K/F is finite separable. Therefore a finite Galois extension K/F gives an étale cover
SpecK→SpecF.
This is why the principal bundle statement is made in the étale topology. In the ordinary Zariski topology, SpecF has only one point, so there are no nontrivial local covers that can trivialize SpecK→SpecF. The étale topology adds algebraic covers like finite separable extensions.
For typing é on macOS:
Option+e, then e
So:
- Press
Option + e
- Release
- Press
e
For uppercase:
Option+e, 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, X is just an arbitrary scheme.
For your Galois-extension example, take
X=SpecF.
Then the étale cover/principal bundle is
SpecK⟶SpecF.
So in this example:
X=SpecF,P=SpecK,G=Gal(K/F).
The map
SpecK→SpecF
comes from the field inclusion
F↪K
by contravariance of Spec. In algebraic geometry, ring maps reverse direction:
F↪K
induces
SpecK→SpecF.
The reason I introduced a general X was to define the étale topology on any scheme. For a general scheme X, objects of Xeˊt are étale maps
In the Galois case, the base scheme is specifically
X=SpecF,
and one distinguished étale object over it is
SpecK→SpecF.
If K/F is finite separable, this map is finite étale. If K/F is finite Galois with group G, it is a principal G-bundle in the étale topology.
User
axiomatically define scheme and give examples.
Assistant
A scheme is a space locally modeled on commutative rings.
Fix CRing = category of commutative unital rings.
A locally ringed space is a pair
(X,OX)
where:
Topological space: X is a topological space.
Structure sheaf: OX is a sheaf of commutative unital rings on X.
Local stalks: For every x∈X, the stalk
OX,x
is a local ring, i.e. it has a unique maximal ideal.
The sheaf OX is the “ring of functions” on X, though its elements may be algebraic rather than ordinary functions.
Given A∈CRing, define
SpecA
as follows.
Points:
∣SpecA∣={p⊆A:p is a prime ideal}.
Closed sets: For each ideal I⊆A,
V(I)={p∈SpecA:I⊆p}.
These define the Zariski topology.
Basic open sets: For f∈A,
D(f)={p∈SpecA:f∈/p}.
Structure sheaf: Define
OSpecA(D(f))=Af,
where Af is the localization of A obtained by inverting f. These assignments glue to a sheaf OSpecA.
Stalks: At p∈SpecA,
OSpecA,p≅Ap.
Since Ap is local, SpecA is a locally ringed space.
An affine scheme is a locally ringed space isomorphic to
(SpecA,OSpecA)
for some A∈CRing.
A scheme is a locally ringed space
(X,OX)
such that:
Affine local cover:
X=i∈I⋃Ui
for open subsets Ui⊆X, and for each i, there exists a commutative unital ring Ai such that
(Ui,OX∣Ui)≅SpecAi
as locally ringed spaces.
So:
scheme=locally ringed space locally isomorphic to SpecA.
Examples
If F is a field, then
SpecF={(0)}.
It has one point, and its structure sheaf has stalk
OSpecF,(0)=F.
This is the scheme appearing in the Galois-extension example:
SpecK→SpecF.
For a field k,
Ak1:=Speck[x].
Its points are prime ideals of k[x]. For example, if k is algebraically closed, the closed points are
for a∈k, plus the generic point (0).
Akn:=Speck[x1,…,xn].
This is the scheme version of n-dimensional affine space over k.
SpecZ.
Its points are
(0), (2), (3), (5), (7),…
where (p) ranges over prime ideals generated by rational primes p.
The point (0) is the generic point. The points (p) are closed.
Projective n-space over k is
It is not usually affine, but it is a scheme because it has an affine open cover
Pkn=i=0⋃nUi,
with
Ui≅Akn.
So schemes include both affine objects like SpecA and objects glued from affine pieces, like Pkn.