port Nullstellensatz & Varieties notes (ch5) (#6)

* 5.1, Corollary 5.2

* variety def, Nullstellensatz proof, correspondences V - I

* port notes on propositions 5.3, 5.7, 5.8

* add missing conclusion at proposition 5.3
This commit is contained in:
2026-01-25 22:04:47 +01:00
committed by GitHub
parent 38228a34ff
commit 5e08a390b9
3 changed files with 323 additions and 2 deletions

View File

@@ -99,7 +99,7 @@
\end{defn}
\begin{defn}{}[prime ideal]
if $a, b \in R$ with $ab \in P$ and $P \neq R$ ($P$ a prime ideal), implies $a in P$ or $b \in P$.
if $a, b \in R$ with $ab \in P$ and $P \neq R$ ($P$ a prime ideal), implies $a \in P$ or $b \in P$.
\end{defn}
\begin{defn}{}[principal ideal]
@@ -1338,7 +1338,7 @@ Recall: a $K$-algebra $A$ is fingen over $K$ if $A=K[y_1, \ldots, y_n]$ for some
thus there exists inverse in $A$, so $A$ is a field too.
\end{proof}
\begin{thm}{R.4.10}[Weak Nullstellensatz - Zariski's lemma]
\begin{thm}{R.4.10}[Weak Nullstellensatz - Zariski's lemma] \label{zariski}
let $k$ a field, $K$ a $k$-algebra which
\begin{enumerate}
\item is finitely generated as a $k$-algebra
@@ -1372,7 +1372,326 @@ Recall: a $K$-algebra $A$ is fingen over $K$ if $A=K[y_1, \ldots, y_n]$ for some
\end{proof}
\vspace{1cm}
\section{Nullstellensatz}
Note: for $k$ a field, $k[X_1, \ldots, X_n]$, $m$ maximal ideal; the residue field $K=k[X_1, \ldots, X_n]/m$ satisfies the Zariski's lemma (\ref{zariski}), thus $K$ is a finite algebraic extension of $k$.
\vspace{0.3cm}
\begin{cor}{5.2} \label{5.2}
$k$ algebraically closed. Then every maximal ideal of $A = k[X_1, \ldots,
X_n]$ is of the form
$$m = (X_1 - a_1, \ldots, X_n -a_n),~~ a_i \in k$$
The map $k[X_1, \ldots, X_n] \longrightarrow k[X_1, \ldots, X_n]/m=k$ is the natural evaluation map $f(X_1, \ldots, X_n) \longmapsto f(a_1, \ldots, a_n)$.
Thus
\begin{align*}
k^n &\longleftrightarrow m-Spec A\\
(a_1, \ldots, a_n) &\longleftrightarrow f(a_1, \ldots, a_n)
\end{align*}
\end{cor}
\begin{proof}
let $m \subset k[X_1, \ldots, X_n]$ be a maximal ideal.
By fundamental property of maximal ideals, $K=A/m$ is a field.
Since $A$ is a fingen $k$-algebra (generated by $X_1, \ldots, X_n$), then $K=A/m$ is also a fingen $k$-algebra, generated by residues $x_i' = x_i +m$.
By Zariski's lemma (\ref{zariski}), $K=A/m$ is algeraic over $k$.
Since by hypothesis $k$ is algebraically closed, it has no proper algebraic extensions\\
\hspace*{2em} $\Longrightarrow~~ K=k~~ \Longrightarrow~~ k \cong A/m$.
So, $\forall x_i \in k$, its image in the quotient field $A/m$ must be an element of $k$.
$$\Longrightarrow x_i'=a_i \in k, ~\forall i \in [n]$$
$$\Longrightarrow x_i - a_i \in m$$
The ideal generated by these terms is a subset of $m$:
$$J=(X_1 -a_1, \ldots, X_n -a_n) \subseteq m$$
Since $J$ is the kernetl of the evaluation map at point $(a_1, \ldots, a_n)$,
then $J$ is a maximal ideal. Together with $J \subseteq m$, then we have
$J=m$, ie. $$m = (X_1 -a_1, \ldots, X_n -a_n)$$
\vspace{0.3cm}
Let
\begin{align*}
\psi: k[X_1, \ldots, X_n] &\longrightarrow k[X_1, \ldots, X_n]/m\\
\psi: x_i &\longmapsto a_i
\end{align*}
Since $\psi$ is a $k$-algebra homomorphism, then $\forall f \in A$:
$$\psi(f(X_1, \ldots, X_n)) = f(\psi(x_1), \ldots, \psi(x_n))= f(a_1, \ldots, a_n)$$
Thus there is a one-to-one correspondence:
points in $k^n ~~~ \longleftrightarrow~~~ m-Spec A$ (maximal ideals in $k[X_1, \ldots, X_n]$
$(a_1, \ldots, a_n) ~~~\longleftrightarrow~~~ (X_1 - a_1, \ldots, X_n - a_n)$
\end{proof}
\vspace{0.5cm}
\subsection{Variety}
\begin{defn}{5.3}[Variety]
A \emph{variety} $V \subset k^n$:
$$ V = V(J) = \{ P=(a_1, \ldots, a_n) \in k^n | f(P)=0 ~\forall~ f \in J \}$$
\begin{itemize}
\item[$\rightarrow$] $V$ is defined by $f_1(P)= \ldots = f_m(P) = 0$
\item[$\rightarrow$] $V$ is defined as the simultaneous solutions of a number of polynomial equations.
\end{itemize}
\end{defn}
\begin{prop}{5.3} \label{5.3}
$k$ an algebraically closed field, and $A=k[X_1, \ldots, X_n]$ a fingen $k$-algebra of the form $A=k[X_1, \ldots, X_n]/J$, where $J$ is an ideal of $k[X_1, \ldots, X_n]$.
(notation: $x_i = X_i \pmod J$)
Then every maximal ideal of $A$ is of the form
$$(x_1 -a_1, \ldots, x_n - a_n)$$
for some point $(a_1, \ldots, a_n) \in V(J)$.
Therefore, $\exists$ a one-to-one correspondence
\begin{align*}
V(X) &\longleftrightarrow m-Spec A\\
\text{given by}~~~(a_1, \ldots, a_n) &\longleftrightarrow (x_1 -a_1, \ldots, x_n - a_n)
\end{align*}
\end{prop}
\begin{proof}
the ideals of $A$ are given by ideals of $k[X_1, \ldots, X_n]$ containing $J$, since for $Q=R/I$, $\exists$ one-to-one correspondence between ideals of $Q$ and ideals of $R$ that contain $I$, ie.
\begin{align*}
m \subset A \longleftrightarrow &m' \subseteq k[X_1, \ldots, X_n]\\
&\text{s.th.}~ J \subseteq m'
\end{align*}
Thus, every maximal ideal of $A$ is of the form
$$\underbrace{(x_1 -a_1, \ldots, x_n - a_n)}_{i.} ~\text{such that}~ \underbrace{J \subset (X_1 - a_1, \ldots, X_n - a_n)}_{ii.}$$
\begin{enumerate}[i.]
\item Since $k$ is algebraically closed, the maximal ideals of $k[X_1, \ldots, X_n]$ look like $m=(X_1 - a_1, \ldots, X_n - a_n)$ for some point $(a_1, \ldots, a_n) \in k^n$.
Which when projected to the quotient ring $A$, $X_i \longmapsto x_i$ (residue class), giving $(x_1 - a_1, \ldots, x_n - a_n)$.
\item for $m$ to exist in $A$, the corresponding $m'$ must contain $J$; since if it didn't contain $J$ it wouldn't "survive" the quotient process.
\end{enumerate}
\vspace{0.3cm}
However, since $(X_1 - a_1, \ldots, X_n - a_n)$ is the kernel of the evaluation map $f \longmapsto f(a_1, \ldots, a_n)$
$\Longrightarrow~$ means that $m'=(X_1 - a_1, \ldots, X_n - a_n)$ consists of all polynomials that vanish at point $P=(a_1, \ldots, a_n)$.
If $J \subseteq m'$, then $\forall~ f \in J$ must vanish at $P$.
By definition, the set of points where all polynomials in $J$ vanish is the \emph{variety}, $V(J)$.
\vspace{0.4cm}
Thus,\\
every maximal ideal in $A$ corresponds to a point $(a_1, \ldots, a_n) \in k^n$, ie.
$$m-Spec A \longleftrightarrow k^n$$
The condition that the ideal belongs to the quotient ring $A=k[X_1, \ldots, X_n]/J$ forces that point to lie in $V(J)$, so
\begin{align*}
m-Spec A &\longleftrightarrow V(J)\\
\text{maximal spectrum} &\longleftrightarrow \text{variety}
\end{align*}
\end{proof}
\begin{prop}{5.5}[Correspondeces $V$ and $I$] \label{5.5}
A variety $X \subset k^n$ is by definition $X=V(J)$ (J an ideal of $k[X_1, \ldots, X_n]$).
So $V$ gives a map:\\
$$\{ \text{ideals of}~ k[X_1, \ldots, X_n] \} \stackrel{V}{\longrightarrow} \{ \text{subsets}~ X ~\text{of}~ k^n \}$$
correspondence going the other way:
$$\{ \text{subsets}~ X ~\text{of}~ k^n \} \stackrel{I}{\longrightarrow} \{ \text{ideals of} ~k[X_1, \ldots, X_n] \}$$
defined by taking a subset $X \subset k^n$ into the ideal
$$I(X) = \{ f \in k[X_1, \ldots, X_n] | f(P)=0 ~\forall~ P \in X \}$$
\vspace{0.4cm}
$V,~I$ satisfy reverse inclusions:
$$J \subset J' ~\Longrightarrow~ V(J) \supset V(J') ~~~~\text{and}~~~~ X \subset Y ~\Longrightarrow~ I(X) \supset I(Y)$$
\end{prop}
\vspace{0.4cm}
\subsection{Nullstellensatz}
\vspace{0.5cm}
\begin{thm}{5.6}[Nullstellensatz] \label{nullstellensatz}
Let $k$ algebraically closed field.
\begin{enumerate}[a.]
\item if $J \subsetneq k[X_1, \ldots, X_n]$ then $V(J) \neq \emptyset$
\item $I(V(J)) = rad J$, in other words, for $f \in k[X_1, \ldots, X_n]$,
$$f(P)=0 ~\forall~ P \in V ~~\Longleftrightarrow~ f^n \in J ~\text{for some $n$.}$$
\end{enumerate}
\end{thm}
\begin{proof}
\begin{enumerate}[a.]
\item if $J \subsetneq k[X_1, \ldots, X_n]$ then $V(J) \neq \emptyset$:\\
Let $m \subset k[X_1, \ldots, X_n]$ be a maximal ideal.\\
Then $L=k[X_1, \ldots, X_n]/m$ is a field (by TODO ref).
By Zariski's lemma (\ref{zariski}), since $L$ is generated as a $k$-algebra by the images of the variables $x_i$, and $k$ is algebraically closed.
Then the only algebraic extension of $k$ is $k$ itself. Thus $L \cong k$.
\vspace{0.3cm}
Then $\exists$ a surjective homomorphism $\psi: k[X_1, \ldots, X_n] \longrightarrow k$.
Let $a_i = \psi(x_i)$. Then $x_i - a \in ker(\psi) = m ~\forall~ i$.
Since the ideal $(X_1 - a_1, \ldots, X_n - a_n)$ is maximal and contained in $m$, they must be equal, ie. $m = (X_1 - a_1, \ldots, X_n - a_n)$.
Therefore, $P=(a_1, \ldots, a_n) \in k^n$ is a zero for every polynomial in $m$.
Since $J \subseteq m$, $P$ is also a zero for every polynomial in $J$.\\
$\Longrightarrow~$ thus $P \in V(J)$, and thus $V(J) \neq \emptyset$.
\item $I(V(J)) = rad J$:\\
\begin{align*}
I(V(J)) &= rad J\\
\text{vanishing ideal of a variety} &= \text{radical of the ideal defining the variety}
\end{align*}
where $rad~J = \{ f \in R ~|~ f^n \in J ~\text{for some}~ n>0 \}$.
Want to show that if a polynomial vanishes at all points where $g_1, \ldots, g_m$ vanish, then $f \in rad(g_1, \ldots, g_m)$.
Consider the ring $k[X_1, \ldots, X_n, Y]$ and the ideal $J'$ generated by $\{ g_1, \ldots, g_m, 1-Y f \}$
Suppose there is a point $(a_1, \ldots, a_n, a_{n+1})$ that is a zero of $J'$. ie.
$$\exists~ (a_1, \ldots, a_n, a_{n+1}) \in V(J')$$
Since $g_i(a)=0$, our hypothesis says $f(a)=0$. However, the last generator $(1-Yf)$ requires
$$1 - a_{n+1} f(a) = 0 ~~\Longrightarrow~ \text{implies}~ 1 - a_{n+1} \cdot 0 = 0 ~\Longrightarrow~ 1-0=0$$
a contradiction.
Therefore, $V(J') = \emptyset$.
\vspace{0.4cm}
Since $V(J')=\emptyset$, by the Weak Nullstellensatz/Zariski (\ref{zariski}),\\
\hspace*{2em} if $V(J')=\emptyset$ then $J'=(1)$, so $1 \in J'=(1)$.
Every element in an ideal is a linear combination of its generators: $J'$ is generated by $\{ g_1, \ldots, g_m, 1-Yf \}$
$$\Longrightarrow~ \forall j \in J',~~ j=(\sum \text{(polynomial)} g_i) + \text{(polynomial)} \cdot (1 - Yf)$$
which, since $1 \in J'$,
$$1= \left( \sum_{i=1}^m p_i(X,Y) g_i(X) \right) + q(X,Y) \cdot (1 - Y f(X))$$
substitute $Y=1/f$,
$$1= \left( \sum_{i=1}^m p_i(X,\frac{1}{f}) g_i(X) \right) + q(X,\frac{1}{f}) \cdot \underbrace{(1 - \frac{1}{f} f(X))}_{0}$$
thus
$$1= \sum_{i=1}^m p_i(X,\frac{1}{f}) g_i(X)$$
multiply by $f^n$,
$$f^n= \sum_{i=1}^m A_i(X) g_i(X)$$
thus $f^n$ is a linear combination of $g_i$.\\
Thus $f^n \in J$, so $f \in rad~J$.
\end{enumerate}
\end{proof}
\vspace{0.4cm}
\subsection{Irreducible varieties}
\begin{defn}{5.7}[Irreducible variety]
a variety $X \subset k^n$ is \emph{irreducible} if it is nonempty and not the union of two proper subvarieties; that is, if
$$X = X_1 \cup X_2 ~~\text{for varieties}~ X_1, X_2 ~\Longrightarrow~ X= X_1 ~\text{or}~ X_2$$
\end{defn}
\begin{prop}{5.7} \label{5.7}
a variety $X$ is irreducible iff $I(X)$ is prime.
\end{prop}
\begin{proof}
set $I=I(X)$.
if $I$ not prime, then $f,g \in A \setminus I$ be such that $fg \in I$.
Define new ideals
$$J_1=(I, f) ~~\text{and}~~ J_2=(I, g)$$
Then, since $f \not\in I(X)$, it follows that $V(J_1) \subsetneq X$
\hspace*{2em}$\Longrightarrow$ so $X=V(J_1) \cup V(J_2)$ is reducible.
The converse is similar.
\end{proof}
\begin{cor}{5.8} \label{cor.5.8}
let $k$ algebraically closed field. Then $V$ and $I$ induce one-to-one correspondences
\begin{align*}
\{ \text{radical ideals}~J~\text{of}~k[X_1, \ldots, X_n] \} &\longleftrightarrow \{ \text{varieties}~ X \subset k^n \}\\
\{ \text{prime ideals}~P~\text{of}~k[X_1, \ldots, X_n] \} &\longleftrightarrow \{ \text{irreducible varieties}~ X \subset k^n \}
\end{align*}
Therefore,
$$Spec k[X_1, \ldots, X_n] = \{ \text{irreducible varieties}~ X \subset k^n \}$$
\end{corollary}
\end{cor}
\begin{prop}{5.8} \label{5.8}
let $A= k[x_1, \ldots, x_n]$ a fingen $k$-algebra ($k$ an algebraically closed field).
Write $J$ for the ideal of relations holding between $x_1, \ldots, x_n$, so that $A=k[X_1, \ldots, X_n]/J$.
Then there is a one-to-one correspondence
$$Spec A \longleftrightarrow \{ \text{irreducible subvarieties}~ X \subset V(J) \}$$
\end{prop}
\begin{proof}
By definition, $Spec A = \{ P ~|~ P \subset A ~\text{is prime ideal} \}$.
By Corollary \ref{cor.5.8}:
$$\{ \text{prime ideals}~P~\text{of}~k[X_1, \ldots, X_n] \} &\longleftrightarrow \{ \text{irreducible varieties}~ X \subset k^n \}$$
About varieties:
\begin{itemize}
\item $I(X)$ in $R=k[X_1, \ldots, X_n]$ is the ideal of the variety $X$\\
\hspace*{2em}ie. the set of all polynomials that vanish on every point of $X$.
\item $I(X)$ in $A=k[X_1, \ldots, X_n]/J$, we're not looking at all possible polynomials but at the residue classes.
\end{itemize}
If $P$ is prime in $A~~\Longrightarrow~$ it must correspond to some prime ideal $\mathfrak{P}$ in $R$ (also $J \subset R$).
Then from Nullstellensatz (\ref{nullstellensatz}), every prime ideal $\mathfrak{P}$ in $R$ is the ideal of some irreducible variety $X$.\\
\hspace*{2em}$\Longrightarrow~~ \mathfrak{P}=I(X)$.
Since we're restricted to the ring $A=k[X_1, \ldots, X_n]/J$, the ideal $P$ are the elements of $I(X)$ viewed through the lens of the quotient\\
$$\Longrightarrow~~ P = \{ f+J ~|~ f \in I(X) \}~~ = I(X) ~\text{mod}~J$$
Now, $J$ is the set of equations defining our "universe" $V(J)$.
Since $A=R/J ~~ \Longrightarrow~$ we thus have $J \subseteq R$.
Also we have a correspondence between $A=R/J$ and $R$.
Let $\mathfrak{P}$ be the preimage of $P$ in $R=k[X_1, \ldots, X_n]$, ie.
\begin{align*}
R &\longrightarrow A=R/J\\
\mathfrak{P} &\longmapsto P\\
I(X) &\longmapsto I(X) ~\text{mod}~J
\end{align*}
$\Longrightarrow~ \text{thus}~ J \subseteq \mathfrak{P}$.
Now, $J \susbset \mathfrak{P} = I(X)$;\\
by \ref{5.5}, the set of points where $\mathfrak{P}$ vanishes must be inside the set of points where $J$ vanishes, so
$$V(J) \supseteq V(\mathfrak{P}) = V(I(X)) = X$$
$\Longrightarrow~~ X \subseteq V(J)$, ie. the irreducible variety $X$ must be a subvariety of $V(J)$.
Therefore,
$$\mathfrak{P} \in Spec A \longleftrightarrow X \subseteq V(J)$$
where $X = V(\mathfrak{P})$.
\end{proof}
\newpage