Past exam of the mathematics course of the University of Cambridge / 2026 / iii / Paper 113 (Algebraic Geometry)

Let $R$ be a valuation ring with fraction field $K$. For every commutative squarethe valuative criterion for separatedness says that a finite type morphism $f$ between Noetherian schemes is separated exactly when there is at most one dotted lift $\mathrm{Spec}R\to X$ completing the diagram.

Under the same finiteness hypotheses, the valuative criterion for properness says that $f$ is proper exactly when every such square has a unique lift. Thus separatedness supplies uniqueness, while properness supplies existence as well.

Solved by gpt-5.6-sol high.

The composite $g\circ f$ is proper, hence separated. If two lifts $\mathrm{Spec}R\to X$ solve a valuation-ring lifting problem for $f$, they also solve the corresponding problem for $g\circ f$. The valuative criterion for separatedness for $g\circ f$ makes them equal, so $f$ is separated.

It remains to prove existence. Start with a squareAfter composing the lower map with $g$, properness of $g\circ f$ gives a lift $a:\mathrm{Spec}R\to X$ over $Z$ whose generic restriction is $u$. The two maps $f\circ a$ and $v$ from $\mathrm{Spec}R$ to $Y$ agree on $\mathrm{Spec}K$ and have the same composite with $g$. Since $g$ is separated, its valuative uniqueness criterion gives $f\circ a=v$. Hence $a$ is the required lift for $f$.

The morphism $f$ is of finite type by hypothesis, and it is separated and satisfies valuative existence. The valuative criterion for properness therefore proves that $f$ is proper.

Solved by gpt-5.6-sol high.

Let $R$ be a discrete valuation ring with uniformizer $\pi$, fraction field $K$, and normalized discrete valuation $\nu :K^{\times }\to \mathbb{Z}$. A map $\mathrm{Spec}K\to {\mathbb{P}}_{\mathbb{Z}}^n$ is a projective pointwith at least one nonzero coordinate. Put $m={\min }_{a_i\ne 0}\nu (a_i)$ and set $b_i={\pi }^{-m}{a}_i$. Then every $b_i$ lies in $R$, and at least one $b_j$ is a unit.

On the standard affine chart $U_j$ of projective space, the ratios $b_i/b_j$ all lie in $R$. They therefore define a map $\mathrm{Spec}R\to U_j\subseteq {\mathbb{P}}_{\mathbb{Z}}^n$ whose generic restriction is the original point. This proves existence in the DVR case. The assumed separatedness of ${\mathbb{P}}_{\mathbb{Z}}^n\to \mathrm{Spec}\mathbb{Z}$, through the valuative criterion for separatedness, gives uniqueness. Thus the morphism satisfies the requested DVR form of the valuative criterion for properness.

Solved by gpt-5.6-sol high.

The tensor product of sheaves $\mathcal{F}{\otimes }_{{\mathcal{O}}_X}\mathcal{G}$ is the sheafification of the presheafEquivalently, its stalk at $x$ is ${\mathcal{F}}_x{\otimes }_{{\mathcal{O}}_{X,x}}{\mathcal{G}}_x$.

The direct image sheaf is defined on each open set $V\subseteq Y$ byThe pullback of a sheaf of modules iswhere $f^{-1}\mathcal{E}$ is the inverse image sheaf and the tensor product uses the structural morphism $f^{-1}{\mathcal{O}}_Y\to {\mathcal{O}}_X$ of the given ringed-space morphism.

Solved by gpt-5.6-sol high.

The ${\mathcal{O}}_Y$-module $\mathcal{E}$ is a locally free sheaf of finite rank if every $y\in Y$ has an open neighborhood $V$ and a finite integer $r$ for whichThe rank is locally constant and is therefore constant on each connected component.

Solved by gpt-5.6-sol high.

The adjunction morphism $f^{-1}{f}_{\ast }\mathcal{F}\to \mathcal{F}$, together with $f^{-1}\mathcal{E}\to f^{\ast }\mathcal{E}$, givesAdjunction between the inverse image sheaf and direct image sheaf turns this into the projection formula for sheaves morphismOn local sections it sends a pure tensor $s\otimes e$ over $V\subseteq Y$ to $s\otimes f^{\#}e$ over $f^{-1}V$.

Whether this morphism is an isomorphism is local on $Y$. If $\mathcal{E}{|}_V\cong {\mathcal{O}}_V^{\oplus r}$ for finite $r$, its restriction becomes the canonical identificationThus the projection-formula morphism is an isomorphism whenever $\mathcal{E}$ is locally free of finite rank.

Solved by gpt-5.6-sol high.

For an indexed open cover $\mathcal{U}=(U_i{)}_{i\in I}$ and a sheaf $\mathcal{F}$, the Čech cohomology cochain groups areThe differential is the alternating sum of restrictions:Since $d^2=0$, the cohomologyis well defined.

Solved by gpt-5.6-sol high.

Write $A=k[x,y]$ and $\mathfrak{m}=(x,y)$. Since $f(0,0)\ne 0$, the principal open subscheme $D(f)=\mathrm{Spec}{A}_f$ contains $\mathfrak{m}$, andCover $V$ by the two affine opens $D(xf)$ and $D(yf)$, whose intersection is $D(xyf)$. The degree-zero part of the resulting Čech cohomology complex givesinside the fraction field of $A$. The last equality follows because $A_f$ is a unique factorization domain and a rational function regular after localizing at both $x$ and $y$ has no possible prime factor left in its denominator.

The same affine cover is acyclic, so its degree-one Čech group computes sheaf cohomology and givesBefore localizing at $f$, the quotienthas the $k$-basisWriting $f=c+h$ with $c=f(0,0)\in k^{\times }$ and $h\in (x,y)$, multiplication by $h$ is locally nilpotent on $Q$: for each negative monomial, a sufficiently high power of $(x,y)$ moves every term into $A_x+A_y$. Hence $c+h$ acts invertibly on $Q$ by a finite geometric series on each element. Localizing at $f$ therefore leaves $Q$ unchanged, andThe displayed infinite basis proves that this vector space is infinite-dimensional.

Solved by gpt-5.6-sol high.

For the cover $\mathcal{U}=\{X_1,X_2\}$ of the affine plane with doubled origin, the overlap is the punctured affine plane $U$. Since $H^0(U,{\mathcal{O}}_U)=k[x,y]$, the Čech complex beginsThis map is surjective, and the normalized complex has no terms in degrees at least two. Consequently

The Mayer-Vietoris sequence for sheaf cohomology also gives $H^1(Y,{\mathcal{O}}_Y)=0$, but its next part givesPart b with $f=1$ shows that the group on the right is infinite-dimensional. Thus

This does not contradict the acyclic cover theorem. Although $X_1$ and $X_2$ are affine, their intersection $U$ is not acyclic: it has nonzero first structure-sheaf cohomology. Equivalently, this affine cover does not satisfy the theorem's hypotheses; the doubled-origin plane is not a semi-separated scheme.

Solved by gpt-5.6-sol high.

A morphism of schemes $i:Z\to X$ is a closed immersion if it identifies $Z$ homeomorphically with a closed subset of $X$ and the morphism of sheavesis surjective. Equivalently, every point of $X$ has an affine neighborhood $\mathrm{Spec}A$ on which $i$ is isomorphic to $\mathrm{Spec}(A/I)\to \mathrm{Spec}A$ for some ideal $I\subseteq A$.

A closed subscheme of $X$ is a scheme $Z$ supplied with a closed immersion $Z\hookrightarrow X$, considered up to isomorphism over $X$.

Solved by gpt-5.6-sol high.

On an affine open $W=\mathrm{Spec}A\subseteq X$, write $Z\cap W=V(I)$. Give $Z\cap W$ the reduced induced subscheme structure $\mathrm{Spec}(A/\sqrt{I})$. These constructions agree under localization and therefore glue to a reduced closed subscheme $Z\hookrightarrow X$.

Let $Z^{\prime }\hookrightarrow X$ be another closed immersion with the same underlying closed set. Affine-locally write $Z^{\prime }=\mathrm{Spec}(A/J)$. Since $V(J)=Z\cap W$, its vanishing ideal is $\sqrt{J}=\sqrt{I}$, and the inclusion $J\subseteq \sqrt{J}$ induces a quotient homomorphismContravariance of the spectrum of a commutative ring gives a factorizationThe quotient maps force these local factorizations to agree on overlaps, so they glue. They are also the only possible maps over $X$, which proves uniqueness.

Solved by gpt-5.6-sol high.

Write $X=\mathrm{Spec}B$, $Y=\mathrm{Spec}A$, and let the morphism correspond to a homomorphism $\phi :A\to B$. Put $I=\ker \phi$ andThe injection $A/I\hookrightarrow B$ gives $X\to Z$, and the quotient $A\to A/I$ gives a closed immersion $Z\hookrightarrow Y$, so $f$ factors through $Z$.

If $f$ factors through another closed subscheme $Z^{\prime }=\mathrm{Spec}(A/J)$, then $J\subseteq \ker \phi =I$. The resulting quotient $A/J\to A/I$ induces the unique factorization $Z\to Z^{\prime }\to Y$. Thus $Z$ is the scheme-theoretic image.

It remains to identify its underlying set. A principal open subscheme $D(a)\subseteq \mathrm{Spec}A$ misses $f(X)$ exactly when $D(\phi (a))$ is empty, equivalently when $\phi (a)$ is a nilpotent element. Hence the ideal of functions vanishing set-theoretically on $f(X)$ has radical $\sqrt{\ker \phi }$. The closure is thereforeas required.

Solved by gpt-5.6-sol high.