Past exam of the mathematics course of the University of Cambridge / 2018 / iii / Paper 101 / 3

A $P$-primary ideal is a proper ideal $Q$ with radical of an ideal $\sqrt{Q}=P$, such that $ab\in Q$ and $a\notin Q$ imply $b^n\in Q$ for some $n\ge 1$. Equivalently, when $\sqrt{Q}=P$, the implication is $ab\in Q$, $b\notin P\Rightarrow a\in Q$.

True for a nonempty finite family. Let $Q=Q_1\cap \cdots \cap Q_m$, with each $Q_j$ a $P$-primary ideal. This intersection of ideals is proper. Its radical of an ideal is $P$: if $r\in P$, choose $n_j$ with $r^{n_j}\in Q_j$ and put $N={\max }_j{n}_j$. Then $r^N\in Q$. Conversely, $r^N\in Q$ implies $r\in \sqrt{Q_j}=P$.

If $ab\in Q$ and $b\notin P$, the primary ideal property gives $a\in Q_j$ for every $j$, so $a\in Q$. ThereforeAs usual, the statement about finite intersections excludes the empty intersection, which is the whole ring and is not a primary ideal.

False. In the polynomial ring $R=k[x,y]$ over a field, takeThe ideal $P$ is a maximal ideal, with quotient ring $R/P\cong k$. AlsoIn this last ring, an element with nonzero constant term is a unit, and an element with zero constant term is a multiple of $y$ and has square zero. Thus every zero divisor is a nilpotent element. This is equivalent to $I$ being a primary ideal: a zero product in the quotient ring with one nonzero factor forces the other factor to be nilpotent. Its radical of an ideal is $P$, so $I$ is $P$-primary.

However, $y\notin I$, so $I\ne P$. Moreover $x\in I$ while $x\notin P^2$, since every polynomial in $P^2$ has no term of total degree one. For every $n\ge 2$, $P^n\subseteq P^2$, so $I\ne P^n$. Hence

True, without a Noetherian hypothesis. By definition, the symbolic power is the contraction of an idealwhere the intersection notation means inverse image under $R\to R_P$, even if this map is not injective. In the local ring $A=R_P$, put $\mathfrak{m}=P{R}_P$. Then $P^n{R}_P={\mathfrak{m}}^n$, and $\sqrt{{\mathfrak{m}}^n}=\mathfrak{m}$: every element of $\mathfrak{m}$ has its $n$th power in ${\mathfrak{m}}^n$, whereas a unit cannot have a power in this proper ideal.

The ideal ${\mathfrak{m}}^n$ is $\mathfrak{m}$-primary. Indeed, if $uv\in {\mathfrak{m}}^n$ and $v\notin \mathfrak{m}$, then $v$ is a unit, so $u\in {\mathfrak{m}}^n$. Its contraction of an ideal is therefore $P$-primary: its radical of an ideal contracts to $P$, and the same implication applies to the images of any two elements of $R$. Thus $P^{(n)}$ is always $P$-primary, proving the reverse implication immediately when $P^n=P^{(n)}$.

For the other implication, suppose $P^n$ is $P$-primary. Membership in the contraction can be expressed by clearing denominators:For completeness, if $r/1=a/t$ with $a\in P^n$ and $t\notin P$, equality of fractions gives $u(tr-a)=0$ for some $u\notin P$, so $(ut)r=ua\in P^n$; the converse follows by inverting $s$. Since $s\notin \sqrt{P^n}=P$, the primary ideal property forces $r\in P^n$. The inclusion $P^n\subseteq P^{(n)}$ always holds, giving

True. An irreducible ideal $I$ is a proper ideal such that $I=J\cap K$ forces $I=J$ or $I=K$. Passing to the quotient ring $A=R/I$ turns this into the assertion that zero cannot be the intersection of two nonzero ideals. Also $A$ is Noetherian: ascending chains of ideals lift to ascending chains of ideals containing $I$ in $R$.

Suppose $ab=0$ in $A$, and suppose $b$ is not a nilpotent element. The ascending chain condition makes the chain of annihilatorsstabilize. Choose $n\ge 1$ with $\mathrm{Ann}(b^n)=\mathrm{Ann}(b^{n+1})$. We claimIf $z=ua=v{b}^n$ lies in this intersection, then $bz=uab=0$, so $v{b}^{n+1}=0$. The equality of annihilators implies $v{b}^n=0$, hence $z=0$. Since $b$ is not nilpotent, $(b^n)\ne 0$. Irreducibility of zero therefore forces $(a)=0$, or $a=0$.

We have proved that $ab=0$ and $a\ne 0$ force $b$ to be a nilpotent element. This says precisely that zero is a primary ideal in $A$, and therefore that $I$ is a primary ideal in $R$.

If its radical of an ideal is to be denoted by a prime ideal, this requires no extra hypothesis: for any primary ideal $I$, $uv\in \sqrt{I}$ and $u\notin \sqrt{I}$ imply $u^n{v}^n\in I$ with $u^n\notin I$, so some power of $v$ lies in $I$. Thus $v\in \sqrt{I}$.

False: existence holds, but uniqueness fails. First, the Lasker–Noether theorem supplies existence; here is an argument from the preceding part. If some proper ideal in a Noetherian ring were not a finite intersection of irreducible ideals, the ascending chain condition would let us choose such an ideal $I$ maximal under inclusion. It is not irreducible, so $I=J\cap K$ with both $J$ and $K$ strictly larger than $I$. By maximality, both are finite intersections of irreducible ideals, and therefore so is $I$, a contradiction. Part (d) makes each factor a primary ideal. The whole ring is represented by the empty intersection if that convention is needed.

For a counterexample to uniqueness, work in $R=k[x,y]$ over a field $k$ and put $I=(x^2,xy)$. The Hilbert basis theorem makes $R$ a Noetherian ring. There are two distinct minimal primary decompositions:The ideal $(x)$ is a prime ideal, hence a primary ideal. Each of the other factors has quotient ring isomorphic to $k[x]/(x^2)$, by substituting respectively $y=0$ and $y=x$. As in part (b), each is $(x,y)$-primary.

To check both intersections, let $Q_c=(x^2,y-cx)$, for $c=0,1$. Substitution identifies $R/Q_c$ with $k[x]/(x^2)$. The class of $xg(x,y)$ is zero exactly when $g(0,0)=0$, equivalently when $g\in (x,y)$. Therefore $(Q_c:x)=(x,y)$ andThe two $Q_c$ are different, since $y$ belongs to $Q_0$ but has nonzero class $x$ modulo $Q_1$. Both decompositions are irredundant: $x\notin Q_c$, while $y-cx\notin (x)$. Applying the radical of an ideal operation to the two components in either decomposition gives $(x)$ and $(x,y)$, respectively.

The second uniqueness theorem for primary decomposition explains the distinction: the component at an isolated prime of a primary decomposition is unique, but embedded primary components need not be. The counterexample varies an embedded component.