Skip to main content

3.3 Varieties, dimensions and rational points

Varieties

def: Affine variety

(X,T,OX)RspXAnZc(An):XRspXXV(An)\begin{align*} &\sphericalangle \\ &(X, T, \mathcal O_X) \in \mathcal R^{sp} \\ &\exists X' \in \mathbb A^{n-} \cap \mathcal Z^c(\mathbb A^n): X \cong_{Rsp} X' \\ \hline \\ &X \in \mathcal V(\mathbb A^n) \end{align*}
note

That is affine variety is a ringed space isomorphic to some irreducible closed set of affine space

Proposition 3.3.1: Affine distinguished subsets are varieties

An(X,T,OX)Rsp,c,fOX(X)=A(X)Xf:=XZ(f)XAn+1,Zc(An+1):(X,T,A(X)f)R(Xf,TXf,OXf)\begin{align*} &\sphericalangle \\ &\mathbb A^n \\ &(X, T, \mathcal O_X) \in \mathcal R^{sp, c, -} \\ &f \in \mathcal O_X(X)=A(X) \\ &X_f:=X \setminus Z(f) \\ \hline \\ &\exists X' \in \mathbb A^{n+1,-} \cap \mathcal Z^c(\mathbb A^{n+1}): (X', T', A(X)_f) \cong_R (X_f, T_{|X_f}, \mathcal O_{|X_f}) \end{align*}
note

OXf\mathcal O_{|X_f} is a natural restriction on subsets of XfX_f

Proof

Let ff' be a representative of ff in A(X)A(X). Consider an ideal JRF[x1,,xn,t],J:=I(X),1tfIJ \lhd_R \overline F[x_1, \ldots, x_n, t], J:=\lang I(X), 1-tf'\rang_I. Next, X:=Z(J)={(P,λ),PXf,λ=1f(P)}X':=Z(J)=\{(P, \lambda), P \in X_f, \lambda = \frac{1}{f'(P)}\}. The zero set indeed has this form because on Z(f):1tfZ(f): 1-tf' never goes to 00 and on XfX_f it does go to 00. Also since the last variable t=1/ft=1/f' it's clear that A(X)=A(X)fA(X')=A(X)_f. So let's prove:

(X,T,A(X)f)R(Xf,TXf,OXf)(X', T', A(X)_f) \cong_R (X_f, T_{|X_f}, \mathcal O_{|X_f})

Consider a projection map π:XXf,(P,λ)P\pi: X' \to X_f, (P, \lambda) \mapsto P and inverse π1:XfX,P(P,1f(P))\pi^{-1}: X_f \to X', P \mapsto (P, \frac{1}{f'(P)}). Since A(X)fA(X)_f and OXf(Xf)O_{|X_f}(X_f) are essentially the same (both have polynomial in numerator and some power of ff in denominator), it's easy to see that π\pi is indeed a ringed space isomorphism.

\square

def: Pre-variety

(X,T,OX)Rsp,UiTo:X=i=1kUii:(Ui,OXUi)V(An(i))(X,T,OX)Vpre\begin{align*} &\sphericalangle \\ &(X, T, \mathcal O_X) \in \mathcal R^{sp, -} \\ &\exists U_i \in T_o: X = \bigcup_{i=1}^kU_i \\ &\forall i: (U_i, \mathcal O_{X|U_i}) \in \mathcal V(\mathbb A^{n(i)}) \\ \hline \\ &(X, T, \mathcal O_X) \in \mathcal V^{pre} \end{align*}

Example: Affine pre-variety

Obvously any affine variety is a pre-variety.

Proposition 3.3.2: Projective variety is a pre-variety

Pn(X,T,OX)Rsp,c,(X,T,OX)Vpre\begin{align*} &\sphericalangle \\ &\mathbb P^n \\ &(X, T, \mathcal O_X) \in \mathcal R^{sp, c, -} \\ \hline \\ &(X, T, \mathcal O_X) \in \mathcal V^{pre} \end{align*}

Proof

Consider X0:={[a0,,an],a00}X_0:= \{[a_0, \ldots, a_n], a_0 \ne 0\}. If I(X)=f1,,frII(X)= \lang f_1, \ldots, f_r \rang_I and define gi(x1,,xn):=fi(1,x1,,xn)g_i(x_1, \ldots, x_n):=f_i(1, x_1, \ldots, x_n), Y:=Z(g1,,grI)Y:=Z(\lang g_1, \ldots, g_r \rang_I). It's easy to build a morphism:

ϕ:X0Y,[a0,,an][a1a0,,ana0]ϕ1:YX0,(a1,,an)[1,a1,,an]\phi: X_0 \to Y, [a_0, \ldots, a_n] \to [\frac{a_1}{a_0}, \ldots, \frac{a_n}{a_0}] \\ \phi^{-1}: Y \to X_0, (a_1, \ldots, a_n) \mapsto [1, a_1, \ldots, a_n]

In this case:

ϕ(p(a1,,an)q(a1,,an))=p(a1a0,,ana0)q(a1a0,,ana0)ϕ1(p(a0,,an)q(a1,,a1))=p(1,a1,,an)q(1,a1,,an)\phi^*(\frac{p(a_1, \ldots, a_n)}{q(a_1, \ldots, a_n)})=\frac{p(\frac{a_1}{a_0}, \ldots, \frac{a_n}{a_0})}{q(\frac{a_1}{a_0}, \ldots, \frac{a_n}{a_0})} \\ \phi^{-1*}(\frac{p(a_0, \ldots, a_n)}{q(a_1, \ldots, a_1)})=\frac{p(1, a_1, \ldots, a_n)}{q(1, a_1, \ldots, a_n)}

So these functions are regular and it's easy to see that closed sets goes to closed sets. So

X0RspYX_0 \cong_{Rsp} Y

In the same way we can define isomorphism for XiX_i and we know that X=iXiX = \bigcup_i X_i so XVqX \in \mathcal V_q

\square

def: Variety

(X,T,OX)Vpre(Y,T,OY)Vpre,f1,f2:YRspX    {PY:f1(P)=f2(P)}Tc(X,T,OX)V\begin{align*} &\sphericalangle \\ &(X, T, \mathcal O_X) \in \mathcal V^{pre} \\ &\forall (Y, T', \mathcal O_Y) \in \mathcal V^{pre}, f_1, f_2: Y \rightsquigarrow_{Rsp}X \implies \\ &\{P \in Y: f_1(P)=f_2(P)\} \in T'^c \\ \hline \\ &(X, T, \mathcal O_X) \in \mathcal V \end{align*}
note

We define the general notion of variety for the sake of completness. Affine and projective varieties (meaning irreducible closet sets) are varieties.

def: Non-singular point

SnXV(Sn)I(X)=f1,,fkIPXrank fi/xj(P)=ndimXPS(X)P non-singular point of X\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &X \in \mathcal V(\mathbb S^n) \\ &I(X) = \lang f_1, \ldots, f_k \rang_I \\ &P \in X \\ &\text{rank } || \partial f_i/\partial x_j(P)||=n-\dim X \\ \hline \\ &P \in \cancel S(X) \\ &P - \text{ non-singular point of }X \end{align*}

def: Non-singular variety

SnXV(Sn)X=S(X)Xnon-singular variety\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &X \in \mathcal V(\mathbb S^n) \\ &X=\cancel S(X) \\ \hline \\ &X - \text{non-singular variety} \end{align*}

Dimensions

def: Dimension of a topological space

(X,T)TdimX:=supn{X0X1Xn,XiXTc}\begin{align*} &\sphericalangle \\ &(X, T) \in \mathcal T\\ \hline \\ &\dim X:=\sup_n\{X_0 \subset X_1 \subset \ldots \subset X_n, X_i \in X^- \cap T^c \} \end{align*}

Proposition 3.3.3: Dimension of closed set is a dimension of it's coordinate ring

Sn(X,T)TcdimX=dimA(X)\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &(X, T) \in \mathcal T^{c} \\ \hline \\ &\dim X = \dim A(X) \end{align*}

Proof

By (3.1.13)(3.1.13) each chain of irreducible closed sets X0XnXX_0 \subset \ldots \subset X_n \subseteq X correspond to prime ideals P0PnI(X)P_0 \supset \ldots \supset P_n \supseteq I(X) and by (2.3.12)(2.3.12) these correspond to prime ideals in A(X)A(X). Since these two correspondences are bijections, we finish the proof.

We'll state two propositions from commutative algebra without a proof

Proposition 3.3.4: Krull dimension and transcendence degree

FFBRID,Bfinitely generated F-algebradimB=trdeg F(B)/FpPI(B)    ht p+dimB/p=dim B\begin{align*} &\sphericalangle \\ &F \in \mathcal F \\ &B \in \mathcal R^{\mathcal {ID}}, B - \text{finitely generated F-algebra} \\ \hline \\ &\begin{align*} & \dim B = \text{trdeg }\mathfrak F(B)/F\hspace{0.5cm} \tag{a}\\ & \mathfrak p \in \mathfrak P_I(B) \implies \text{ht } \mathfrak p + \dim B/\mathfrak p = \text{dim } B\hspace{0.5cm} \tag{b}\\ \end{align*} \end{align*}

Proposition 3.3.5: UFD and prime ideals criterion

NRNNRUFD    pPI(N),ht p=1:fN:p=fI\begin{align*} &\sphericalangle \\ &N \in \mathcal R^{\mathcal {N}} \\ \hline \\ &N \in \mathcal R^{\mathcal {UFD}} \iff \forall \mathfrak p \in \mathfrak P_I(N), \text{ht } \mathfrak p = 1: \exists f \in N: \mathfrak p = \lang f \rang_I \end{align*}

Proposition 3.3.6: Krull's hauptidealsatz

NRNfNpPI(N):fp,pPI(N):fppht p=1\begin{align*} &\sphericalangle \\ &N \in \mathcal R^{\mathcal N} \\ &f \in N \\ &\mathfrak p \in \mathfrak P_I(N): f \in \mathfrak p, \nexists \mathfrak p' \in \mathfrak P_I(N): f \in \mathfrak p' \subseteq \mathfrak p \\ \hline \\ &\text{ht }\mathfrak p = 1 \end{align*}

Proposition 3.3.7: Dimension of an open set equals the dimension of its closure

(X,T)Tc,UTodimU=dimU\begin{align*} &\sphericalangle \\ &(X, T) \in \mathcal T^{c,-} \\ &U \in T^o \\ \hline \\ &\dim U = \dim \overline U \end{align*}

Proof

Note that by (3.1.1)(3.1.1) (U,TU)T(U, T_{|U}) \in \mathcal T^- and by (3.1.2)(3.1.2) we have (U,TU)Tc,(\overline U, T_{|\overline U}) \in \mathcal T^{c,-}. Consider a chain Z0Zn,ZiTUc,Z_0 \subset \ldots \subset Z_n, Z_i \in T^{c,-}_{|U}. Now take the closure of these sets in XX to get the chain Z0Zn,ZiTc\overline Z_0 \subseteq \ldots \subseteq \overline Z_n, \overline Z_i \in T^{c}. First, let's prove that Zi=ZiUZ_i=\overline Z_i \cap U:

UZi=UXiZi,XiTcXi=XiZi,XiTc(UXi)=UXiUZi,XiTc(UXi)=XiZi,XiTUcXi=Zi since (ZiTUc)U \cap \overline Z_i = U \cap \bigcap_{X_i \supseteq Z_i, X_i \in T^c} X_i = \\ \bigcap_{X_i \supseteq Z_i, X_i \in T^c} (U \cap X_i) = \bigcap_{U \cap X_i \supseteq U \cap Z_i, X_i \in T^c} (U \cap X_i)= \\ \bigcap_{X_i' \supseteq Z_i, X_i' \in T^c_{|U}} X_i' = Z_i \text{ since } (Z_i \in T^c_{|U})

Next, ZiTUc,\overline Z_i \in T^{c,-}_{|\overline U}. It's closed simply because UX\overline U \subseteq X. If Zi\overline Z_i was reducible, that is Zi=X1X2,X1,2Zi    Zi=UZi=(UX1)(UX2)    (X1U)(X2U)    (X1=X1=U)(X2=X2=U)\overline Z_i = X_1 \cup X_2, X_{1,2} \subset \overline Z_i \implies Z_i=U \cap \overline Z_i = (U \cap X_1) \cup (U \cap X_2) \implies (X_1 \supseteq U) \vee (X_2 \supseteq U)\implies (X_1=\overline X_1 = \overline U) \vee (X_2=\overline X_2 = \overline U). But this is a contradiction since X1,2ZiUX_{1,2} \subset \overline Z_i \subseteq \overline U. So we establised:

ZiTUc,Z_i \in T^{c, -}_{|\overline U}

So we know the chain Z0Zn,ZiTUc,Z_0 \subset \ldots \subset Z_n, Z_i \in T^{c,-}_{|U} implies the existence of chain Z0Zn,ZiTUc,\overline Z_0 \subset \ldots \subset \overline Z_n, \overline Z_i \in T^{c,-}_{|\overline U}. There could not be equivalence \subseteq in this chain as we know that Zi=UZiZ_i = U \cap \overline Z_i so Zi=Zi+1    UZi=UZi+1    Zi=Zi+1\overline Z_i = \overline Z_{i+1} \implies U \cap \overline Z_i = U \cap \overline Z_{i+1} \implies Z_i = Z_{i+1}.

Now we can go the other way round and project some chain X0Xn,XiTUc,X_0 \subset \ldots \subset X_n, X_i \in T^{c,-}_{|\overline U} to UU by taking Xi=UXiX_i'=U \cap X_i which will be closed and irreducible (the latter can be proving by assuming otherwise and taking closure similarly to the above). So we have a bijection of irreducible closed chains. The last thing we need to deal with is what if UXi=UXi+1U \cap X_i = U \cap X_{i+1}? Then we'll have a shorter chain of XiX_i'. If UXi=UXi+1U \cap X_i = U \cap X_{i+1} then Xi+1XiUUX_{i+1} \setminus X_i \in \overline U \setminus U. On the other hand Xi+1=XiXi+1Xi    Xi+1TUc,Xi+1Xi=Xi+1X_{i+1}=X_i \cup \overline {X_{i+1} \setminus X_i} \overset{X_{i+1} \in T^{c,-}_{|\overline U}}\implies \overline {X_{i+1} \setminus X_i} = X_{i+1}. At the same time UUTUc\overline U \setminus U \in T^c_{|\overline U} since UToU \in T^o. So we have Xi+1=Xi+1XiUU=UUX_{i+1}=\overline {X_{i+1} \setminus X_i} \subseteq \overline{\overline U \setminus U}=\overline U \setminus U. But it cannot be since Xi+1UX_{i+1} \cap U \ne \empty.

So we have fully matched chains of irreducible in UU and U\overline U and dimU=dimU\dim U = \dim \overline U

\square

Proposition 3.3.8: Dimension of a variety with a principal vanishing ideal

Sn(X,T)Tc,dimX=n1    fF[x1,,xn]:X=Z(f)\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &(X, T) \in \mathcal T^{c,-}\\ \hline \\ &\dim X = n-1 \iff \exists f \in \overline F[x_1, \ldots, x_n]^-: X = Z(f) \end{align*}

Proof

    \implies

dimX=n1    n1=(3.3.3)dimA(X)=dimF[x1,,xn]/I(X)=(3.3.4.b)dimF[x1,,xn]ht I(X)    ht I(X)=1    (3.3.5)fF[x1,,xn]:I(X)=fI    X=Z(f)\dim X = n-1 \implies n-1 \overset{(3.3.3)}= \dim A(X) = \\ \dim \overline F[x_1, \ldots, x_n] / I(X) \overset{(3.3.4.b)}= \\ \dim \overline F[x_1, \ldots, x_n] - \text{ht }I(X) \implies \text{ht }I(X) = 1 \overset{(3.3.5)}\implies \\ \exists f \in \overline F[x_1, \ldots, x_n]: I(X) = \lang f \rang_I \implies X = Z(f)\\

    \impliedby

fF[x1,,xn]    I(X)=fIPI(F[x1,,xn])    (3.3.6)ht I(X)=1    (3.3.4.b)dimA(X)=dimF[x1,,xn]/I(X)=n1    (3.3.3)dimX=n1f \in \overline F[x_1, \ldots, x_n]^- \implies I(X) = \lang f \rang_I \in \mathfrak P_I(\overline F[x_1, \ldots, x_n]) \overset{(3.3.6)}\implies \text{ht } I(X) = 1 \overset{(3.3.4.b)}\implies \dim A(X)=\dim \overline F[x_1, \ldots, x_n]/I(X) = n-1 \overset{(3.3.3)}\implies \dim X = n-1

\square

Proposition 3.3.9: Dimension of the local ring

SnXV(Sn)PXmPMI(OX,P)dimOX,Pr=dimX\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &X \in \mathcal V(\mathbb S^n) \\ &P \in X \\ &\mathfrak m_{P} \in \mathfrak M_I(\mathcal O_{X,P})\\ \hline \\ &\dim \mathcal O^r_{X,P} = \dim X \end{align*}

Proof

First let's note that chain of prime ideals in mP\mathfrak m_{P} corresponds to a chain of prime ideals in A(X)mPOX,PrA(X)_{\mathfrak m_{P}} \cong \mathcal O^r_{X,P} so ht mP=dimOX,Pr\text{ht }\mathfrak m_{P}=\dim \mathcal O^r_{X,P}. But from (3.3.4.b)(3.3.4.b) we know that

ht mP+dimA(X)/mP=dimA(X)=dimX\text{ht }\mathfrak m_{P} + \dim A(X)/\mathfrak m_{P} = \dim A(X) = \dim X

Next A(X)/mPFA(X)/\mathfrak m_{P} \in \mathcal F implies dimA(X)/mP=0\dim A(X)/\mathfrak m_{P} = 0 (note we're talking about Krull dimension here) so

dimOX,Pr=ht mP=ht mP+dimA(X)/mP=dimX\dim \mathcal O^r_{X,P}=\text{ht }\mathfrak m_{P}=\text{ht }\mathfrak m_{P} + \dim A(X)/\mathfrak m_{P}=\dim X

\square

Proposition 3.3.10: Non-singular point criterion

SnXV(Sn)PXmPMI(OX,P)PS(X)    dimX=dimFmP/mP2\begin{align*} &\sphericalangle \\ &\mathbb S^n \\ &X \in \mathcal V(\mathbb S^n) \\ &P \in X \\ &\mathfrak m_{P} \in \mathfrak M_I(\mathcal O_{X,P})\\ \hline \\ &P \in \cancel S(X) \iff \dim X=\dim_{\overline F} \mathfrak m_{P}/\mathfrak m^2_{P} \end{align*}
note

Since OX,P/mPFF\mathcal O_{X,P}/\mathfrak m_{P} \cong_F \overline F we know that dimOX,P/mPmP/mP2=dimFmP/mP2\dim_{\mathcal O_{X,P}/\mathfrak m_{P}} \mathfrak m_{P}/\mathfrak m^2_{P}=\dim_{\overline F} \mathfrak m_{P}/\mathfrak m^2_{P}

Proof

We'll make a proof for An\mathbb A^n. Assume P=(a1,,an)AnP=(a_1, \ldots, a_n) \in \mathbb A^n then by (3.1.8)(3.1.8) we have a maximal ideal aP=xa1,,xanI\mathfrak a_{P}=\lang x-a_1, \ldots, x-a_n \rang_I. Consider a homomorphism:

θP:F[x1,,xn]MFn,f(f/x1(P),,f/xn(P))\theta_P: \overline F[x_1, \ldots, x_n] \rightsquigarrow_M \overline F^n, f \to (\partial f/\partial x_1(P), \ldots, \partial f/\partial x_n(P))

If we consider a polynomal from ap:q(x)=(xai)p(x)\mathfrak a_p: q(x)=(x-a_i)p(x) then q(x)=p(x)+(xai)p(P)    q(P)=p(P)q'(x)=p(x)+(x-a_i)p'(P) \implies q'(P)=p(P). So θP(aP)=Fn\theta_P(\mathfrak a_{P})=\overline F^n and kerθP=aP2\ker \theta_P=\mathfrak a_{P}^2 we have

aP/aP2ϕPMFn \mathfrak a_{P} / \mathfrak a^2_{P} \overset{\phi_P}\cong_M \overline F^n

Next, consider I(X)=f1,,fkII(X)= \lang f_1, \ldots, f_k \rang_I and J:=fi/xj(P)IJ:=\lang \partial f_i/\partial x_j(P) \rang_I. Consider some function figI(X)f_ig \in I(X) then θP(fig)=(,fi(P)g(P)+fi(P)g(P),)=(,fi(P)g(P),)=g(P)Ji\theta_P(f_ig)=(\ldots, f_i'(P)g(P) + f_i(P)g'(P), \ldots)=(\ldots, f_i'(P)g(P), \ldots) = g(P)J_i. So θP(I(X))=J1,,JkM\theta_P(I(X)) = \lang J_1, \ldots, J_k \rang_M and thus rank J=dimFθP(I(X))\text{rank }J=\dim_{\overline F} \theta_P(I(X)).

The image θP(I(X))\theta_P(I(X)) is isomorphic to it's image under natural homomorphism in aP2/aP2\mathfrak a_P^2/\mathfrak a_P^2 which is (I(X)+aP2)/aP2(I(X)+\mathfrak a_P^2)/\mathfrak a_P^2. So rank J=dimFθP(I(X))=dimF(I(X)+aP2)/aP2 \text{rank }J=\dim_{\overline F} \theta_P(I(X)) = \dim_{\overline F} (I(X) + \mathfrak a_{P}^2) / \mathfrak a_{P}^2.

Consider MP=aP/I(X)M_P=\mathfrak a_P/I(X) - the image of aP\mathfrak a_P in A(X)A(X) under natural homomorphism. Then mMP2:xaP:m=(x+I(X))2=x2+I(X)\forall m \in M_P^2: \exists x \in \mathfrak a_P: m = (x+I(X))^2= x^2 + I(X) so MP2=(ap2+I(X))/I(X)M_P^2=(\mathfrak a_p^2+I(X))/I(X). Thus by third isomopshism theorem MP/MP2aP/(ap2+I(X))M_P/M_P^2 \cong \mathfrak a_P/(\mathfrak a_p^2+I(X)). Next note that mMI(OX,P)\mathfrak m \in \mathfrak M_I(\mathcal O_{X, P}) then m=(A(X)MP)1MP,m2=(A(X)MP)1MP2\mathfrak m = (A(X) \setminus M_P)^{-1}M_P, \mathfrak m^2 = (A(X) \setminus M_P)^{-1}M_P^2 (we don't double 1-1 power because MPM_P is prime so any two functions not in MPM_P will still be not in MPM_P and obviously we can take one function to be 11 to get the whole (A(X)MP)1(A(X) \setminus M_P)^{-1}) so

m/m2aP/(I(X)+aP2)\mathfrak m / \mathfrak m^2 \cong \mathfrak a_P / (I(X) + \mathfrak a_P^2)

Finally

dimFm/m2+rank J=dimFaP/(I(X)+aP2)+dimF(I(X)+aP2)/aP2=dimFaP/aP2=n\dim_{\overline F} \mathfrak m / \mathfrak m^2 + \text{rank }J= \\ \dim_{\overline F} \mathfrak a_P / (I(X) + \mathfrak a_P^2) + \dim_{\overline F} (I(X) + \mathfrak a_{P}^2) / \mathfrak a_{P}^2 = \dim_{\overline F} \mathfrak a_{P} / \mathfrak a_{P}^2 = n

So

PS(X)    rank J=ndimX    dimFm/m2=dimXP \in \cancel S(X) \iff \text{rank }J = n - \dim X \iff \\ \dim_{\overline F} \mathfrak m / \mathfrak m^2 = \dim X

\square

Rational points

Starting from this point we'll always assume that FFPF \in \mathcal F^{\mathcal P}.

So far we always dealt with F\overline F but it's natural when we want to consider varieties over FF as well. This is where Galois theory comes very handy.

Recall that if we have a Galois extension E/GalFE/_{Gal}F then we can define FF as F=FixGal(E/F)EF=\text{Fix}_{\text{Gal}(E/F)}E. We'll soon see that many notions in algebraic geometry is Galois invariant, meaning we can quickly see how Galois automorphisms acts on these notions. So this is a good tool for analyzing the structure of varieties over FF.

Notice that we always consider Galois extension F/F\overline F / F since by defintion F/F\overline F /_{\lhd} F and since FFpF \in \mathcal F^p we have F/F\overline F /_{\boxminus} F so by (2.9.15)(2.9.15) F/GalF\overline F /_{\text{Gal}} F.

We'll start by stating very simple object - a point PAnP \in \mathbb A^n. Then Galois automorphism acts trivially on PP:

Pσ=(x1σ,,xnσ),xiσ=σ(xi)P^\sigma = (x_1^\sigma, \ldots, x_n^\sigma), x_i^{\sigma}=\sigma(x_i)

Then we can say that:

AFn={PAn:σGal(F/F)    Pσ=P}={(x1,,xn):xiF}\mathbb A^n_F=\{P \in \mathbb A^n:\sigma \in \text{Gal}(\overline F / F) \implies P^{\sigma}=P\}= \\ \{(x_1, \ldots, x_n): x_i \in F\}

Next if we consider a variety XX then we say that XX is defined over FF if vanishing ideal I(X)=f1,,fnI,fiF[x]I(X)=\lang f_1, \ldots, f_n \rang_I, f_i \in F[x].

note

We will use notation X/FX/F or XV/FX \in \mathcal V/F for variety defined over F.

In this case indeed we can consider XF=ZF(I(X))X_F=Z_F(I(X)) in the field FF and each solution of polynomials in the field FF will be a solution in F\overline F and vice versa. We say that XFX_F is a set of FF-rational points.

Next we can define a vanishing ideal:

I(X)F=I(X)F[x1,,xn]I(X)_F=I(X)\cap F[x_1, \ldots, x_n]

Then we'll have:

X/F    I(X)=I(X)FF[x1,,xn]X/F \iff I(X)=I(X)_F \overline F[x_1, \ldots, x_n]

We can consider how Galois automorphisms act on functions. If fF[x1,,xn]f \in \overline F[x_1, \ldots, x_n] then

f(P)σ=fσ(P)σf(P)^\sigma=f^{\sigma}(P)^\sigma

So if fF[x1,,xn]f \in F[x_1, \ldots, x_n]

f(P)σ=f(Pσ)f(P)^\sigma=f(P^\sigma)

We can summarize that if X/FX/F then

XF=XAFn={PX:σGal(F/F)    Pσ=P}X_F=X \cap \mathbb A^n_F=\\\{P \in X: \sigma \in \text{Gal}(\overline F / F) \implies P^{\sigma}=P \}

Finally we can define coordinate ring:

A(X)F:=F[x1,,xn]/I(X)FA(X)_F:=F[x_1, \ldots, x_n]/I(X)_F

And the function field F(X)F:=F(A(X)F)F(X)_F:=\mathfrak F(A(X)_F)