Past exam of the mathematics course of the University of Cambridge / 2014 / iii / Paper 20 / Section A

Write the comonad as $(G,ϵ,\delta )$ and its category of coalgebras for a comonad as ${\mathcal{E}}^G$. A coalgebra for a comonad is a map $a:A\to GA$ with ${ϵ}_{A}a=1_A$ and ${\delta }_{A}a=Gaa$. A morphism $f:(A,a)\to (B,b)$ satisfies $bf=Gfa$. Let $U:{\mathcal{E}}^G\to \mathcal{E}$ be the forgetful functor and let $R(X)=(GX,{\delta }_X)$ be the cofree coalgebra. The adjunction $U⊣R$ has the explicit correspondenceWe construct the three pieces of the elementary topos structure.

Because $G$ preserves finite limits, each underlying finite limiting cone has a unique coalgebra structure induced by the structures on its vertices. The counit and coassociativity equations can be checked after its jointly monic projections. Thus $U$ creates finite limits and reflects isomorphisms. A morphism of coalgebras is monic exactly when its underlying morphism is monic, by the diagonal criterion using the created pullback.

For exponentials in a coalgebra topos, fix coalgebras $(A,a)$ and $(B,b)$ and put $E=B^A$ in $\mathcal{E}$. On the cofree coalgebra $R(E)$ there is an underlying evaluationThe two mapsuse $G(GE\times A)\cong GGE\times GA$ in the second expression. Transpose them in $\mathcal{E}$ to maps $r,s:GE\rightrightarrows (GB{)}^A$, and then transpose across $U⊣R$ to coalgebra morphisms $\tilde{r},\tilde{s}:R(E)\rightrightarrows R((GB{)}^A)$. Take their equalizer $C$ in ${\mathcal{E}}^G$.

An underlying map $UX\times A\to B$ corresponds to a coalgebra map $h:X\to R(E)$. The equation saying that the original map is a coalgebra morphism is precisely $rh=sh$, since ${\delta }_{E}h=Ghx$ for the structure $x:X\to GX$. By the cofree adjunction, this is equivalent to $\tilde{r}h=\tilde{s}h$, hence to unique factorization through $C$. Thereforenaturally in $X$. This constructs the required exponential object.

For the subobject classifier of a coalgebra topos, let $\top :1\hookrightarrow \Omega$ be the underlying subobject classifier, and let $\kappa :G\Omega \to \Omega$ classify the mono $G\top :G1\cong 1\hookrightarrow G\Omega$. Its cofree transpose is the coalgebra endomorphismDefine ${\Omega }_G$ as the equalizer of $k$ and $1_{R\Omega }$. The transpose of $\top$ factors through this equalizer and gives ${\top }_G:1\to {\Omega }_G$.

Indeed, for a subobject $S\hookrightarrow UX$ classified by $\chi :UX\to \Omega$, the pullback $x^{-1}(GS)$ has characteristic map $\kappa G\chi x$. It is always contained in $S$, by naturality of the counit. Equality holds exactly when $x$ restricts to a coalgebra structure on $S$; its axioms then follow by composing with the monomorphisms $GS\hookrightarrow GX$ and $GGS\hookrightarrow GGX$. Under $U⊣R$, the classifying map becomes $h=G\chi x:X\to R\Omega$. The equality of subobjects is ${ϵ}_{\Omega }h=\kappa h$. Transposing this equality gives $h=kh$, so exactly the coalgebra subobjects correspond to maps $X\to {\Omega }_G$. Their pullback of ${\top }_G$ is the desired subcoalgebra, and uniqueness follows from uniqueness of $\chi$.

Thus we have finite limits, exponentials and a subobject classifier:The construction does not assume that $G$ preserves the underlying exponentials or underlying subobject classifier.

A decidable object in a topos has a decomposition $A\times A={\Delta }_A⨿D_A$, with $D_A$ representing inequality. In the internal logic of a topos, equality on $A$ is decidable. These categorical complements behave well under pullback.

For a subobject $B\hookrightarrow A$, pull back the displayed decomposition along $B\times B\hookrightarrow A\times A$. The diagonal pulls back to ${\Delta }_B$, while $D_A$ pulls back to its complement. Hence every subobject of a decidable object is decidable; the subobject itself need not be complemented in $A$.

For two decidable objects, the diagonals of $A$ and $B$ give four disjoint summands of $(A\times B{)}^2$, according as each coordinate pair is equal or unequal. The both-equal summand is ${\Delta }_{A\times B}$, and the other three give its complement. The terminal object is decidable, so induction gives closure under finite products, including the empty product.

For a family $(A_i{)}_{i\in I}$ of decidable objects with an existing coproduct $A={\coprod }_i{A}_i$, products distribute over this coproduct, givingThe coproduct injections in a topos are disjoint. The diagonal consists of ${\Delta }_{A_i}$ in each $i=j$ summand, and has complementConsequently every existing coproduct of decidable objects is decidable, including the initial object. In a Grothendieck topos all small coproducts exist. No assertion that arbitrary products preserve decidability is used.

Let $\mathcal{Q}$ be the full subcategory of quotients of decidable objects. It is closed under quotients, by composition of epimorphisms, and under small coproducts, by part (ii). It is also closed under subobjects: pull back a decidable cover $D\twoheadrightarrow B$ along $B^{\prime }\hookrightarrow B$. The resulting cover of $B^{\prime }$ has domain a subobject of $D$, hence a decidable object.

The terminal object lies in $\mathcal{Q}$. If $D\twoheadrightarrow B$ and $D^{\prime }\twoheadrightarrow C$ are decidable covers, their product is an epimorphism $D\times D^{\prime }\twoheadrightarrow B\times C$. Part (ii) makes its domain decidable. Thus products, and then equalizers as subobjects of products, remain in $\mathcal{Q}$. The inclusion $I:\mathcal{Q}\hookrightarrow \mathcal{E}$ preserves finite limits.

We construct a coreflective subcategory rather than claim that every object has a decidable cover. For $X\in \mathcal{E}$, let $qX\hookrightarrow X$ be the union of all subobjects of $X$ which lie in $\mathcal{Q}$. The Grothendieck topos is well-powered, so these subobjects form a set. Choose a decidable cover of each and take their coproduct. Its map to $X$ has image $qX$, so $qX$ is itself a quotient of a decidable object. Any map from an object of $\mathcal{Q}$ to $X$ has image in $\mathcal{Q}$, and therefore factors uniquely through $qX$. This givesThe induced idempotent comonad $G=Iq$ on $\mathcal{E}$ preserves finite limits: $q$ is a right adjoint and $I$ preserves those limits. Its counit is the inclusion $qX\hookrightarrow X$, and $q(qX)=qX$.

The coalgebras of this comonad are exactly the objects of $\mathcal{Q}$. A coalgebra structure is a section $X\to qX$ of the monic counit, forcing the counit to be an isomorphism; conversely an object already in $\mathcal{Q}$ has the unique such structure. Thus $\mathcal{Q}\simeq {\mathcal{E}}^G$. Part (i) now givesThis argument proves the required elementary-topos conclusion without presuming a small family of decidable generators for $\mathcal{E}$.

A local operator, also called a Lawvere-Tierney topology, is a map $j:\Omega \to \Omega$ which internally satisfiesThe closure operation of a local operator sends a mono with characteristic map $\chi$ to the subobject classified by $j\chi$. It is inflationary, idempotent and pullback-stable. A mono is j-dense if its closure is its whole codomain, and j-closed if it equals its closure. A j-sheaf $S$ is an object for which restrictionis a bijection for every j-dense mono $A\hookrightarrow B$; requiring only injectivity defines a j-separated object.

Here is a construction underlying the sheaf reflector for a local operator. The closed-subobject classifier ${\Omega }_j=\{p:jp=p\}$ is a j-sheaf: closed subobjects on a dense subobject extend uniquely by taking their closure in the larger object. Powers ${\Omega }_j^X$ are also sheaves, because products of a dense mono with $X$ remain dense. A j-closed subobject of a sheaf is a sheaf: first extend a map into the ambient sheaf, then use density to force its image into the closed subobject.

Close the diagonal of $X$. Its j-closure is an equivalence relation, using preservation of finite meets and pullback-stability to verify transitivity. The effective quotient $X\twoheadrightarrow X_s$ is the separated reflection: every map from $X$ into a separated object identifies that closed diagonal and factors uniquely. For separated $X_s$, the closed-singleton mapis monic. Its j-closed image closure $a_{j}X$ is a sheaf, and $X_s\hookrightarrow a_{j}X$ is dense. Unique extension across that mono, following the separated quotient factorization, provesfor every sheaf $S$. This proves reflectivity. The closure construction is pullback-stable; equivalently, separated quotients and the subsequent dense embeddings commute with the finite limiting comparisons, giving the usual left-exact sheaf reflector.

Finite limits of sheaves are computed in $\mathcal{E}$, because unique extensions can be taken componentwise. If $S$ is a sheaf, $S^T$ is a sheaf for any $T$, by the same product-with-dense-mono argument. Thus sheaf exponentials are the ambient exponentials. Monos between sheaves have j-closed images: their closure is a sheaf, and the dense inclusion into it splits by the extension property, hence is an isomorphism. Therefore ${\Omega }_j$ classifies precisely their subobjects. These observations establish ${\mathbf{sh}}_j(\mathcal{E})$ is a reflective topos.

Now let $u:1\to \Omega$ classify the given subterminal object. Its open local operator and closed local operator areThe Heyting algebra identities verify all local-operator axioms: implication by fixed $u$ preserves meets and is idempotent, while adjoining $u$ preserves meets by distributivity and is idempotent.

For a mono in $B$ with characteristic predicate $p$, closedness for $c(U)$ means $u\vee p=p$, or $u\le p$. Density for $o(U)$ means $(u\Rightarrow p)=\top$, again $u\le p$. Thus the c(U)-closed monos are exactly the o(U)-dense monos.

Both densities together force $u\le p$ and $u\vee p=\top$, hence $p=\top$: the only jointly dense monos are isomorphisms. More explicitly, the meet of these operators is pointwise andso $o(U)\wedge c(U)={\mathrm{id}}_{\Omega }$.

For their join, every mono $A\hookrightarrow B$ factors through the union with the pullback $U_B=U\times B$:The first mono is c(U)-dense, because adjoining $U$ fills its codomain; the second is o(U)-dense, because its image contains $U_B$. Any local operator above both must therefore make every mono dense, since its dense monos are closed under composition. It is the largest operator $p\mapsto \top$. ConsequentlyThese are the complementary open and closed local operators in the ordered lattice of local operators, with order given by pointwise implication.

A first-order signature specifies sorts, function symbols with specified input and output sorts, and relation symbols with specified input sorts. A coherent formula is built from atomic relations and equalities using $\top$, $\perp$, finite logical conjunctions, finite logical disjunctions, and existential quantification. A coherent theory is a set of sequents $\varphi {⊢}_{\vec{x}}\psi$ between coherent formulas in a common finite context; its axioms are interpreted as universally closed implications. Neither general negation nor universal quantification is allowed inside coherent formulas.

One complete presentation of coherent logic consists of the following axiom and rule schemes, together with the theory's sequents. All displayed formulas have compatible sorts and contexts, and bound variables can be renamed.

Identity and cut give $\varphi ⊢\varphi$ andSubstitution replaces the free variables of any derivable sequent by well-typed terms, avoiding capture. Contexts can be enlarged by unused variables, and permuted or renamed.

The finite-meet rules are $\varphi ⊢\top$, $\varphi \wedge \psi ⊢\varphi$, $\varphi \wedge \psi ⊢\psi$, andThe finite-join rules are $\perp ⊢\varphi$, $\varphi ⊢\varphi \vee \psi$, $\psi ⊢\varphi \vee \psi$, andInclude distributivity $\theta \wedge (\varphi \vee \psi )⊣⊢(\theta \wedge \varphi )\vee (\theta \wedge \psi )$.

Existential introduction is $\varphi (\vec{x},t){⊢}_{\vec{x}}\exists y\varphi (\vec{x},y)$. Existential elimination iswhere $y$ is absent from $\psi$. Equivalently, the quantifier is left adjoint to weakening along the context projection. Include the Frobenius rule in coherent logicIt can also be derived from the usual coherent natural-deduction rules.

Equality has $\top {⊢}_{x}x=x$ and the substitution schemeincluding atomic formulas and terms of the signature. Symmetry, transitivity and congruence for all functions and relations follow. These schemes impose no unintended inhabitedness axiom on a sort.

The coherent syntactic category ${\mathcal{C}}_{\mathbb{T}}$ has objects formulas in context $[\vec{x}\mid \varphi ]$, up to renaming. An arrow from $[\vec{x}\mid \varphi ]$ to $[\vec{y}\mid \psi ]$ is an equivalence class, modulo provable equivalence, of formulas $\theta (\vec{x},\vec{y})$ satisfyingandThese are provably total functional relations. The identity is the equality graph restricted by $\varphi$. If $\theta$ and $\rho$ are consecutive arrows, their composite is represented by $\exists \vec{y}(\theta \wedge \rho )$. Equality, cut and existential rules give the category laws.

A coherent category has finite limits, pullback-stable regular-epi/mono image factorizations, and finite unions of subobjects stable under pullback. We verify these structures syntactically. The terminal object is the empty-context truth formula. Products conjoin formulas in disjoint contexts; equalizers add equality of the two output tuples. For a functional relation $\theta$, its image in the target is represented by $\exists \vec{x}\theta$. Its factor onto that image is regular epic: two arrows out of the image agreeing on the source agree by existential elimination, and the same argument with the kernel pair gives the coequalizer property. Frobenius makes these image factorizations stable under pullback.

Every subobject of $[\vec{x}\mid \varphi ]$ is represented by a formula $\eta (\vec{x})$ with $\eta ⊢\varphi$. Indeed, take the existential image of a monic functional relation; uniqueness makes its map to that image an isomorphism. Subobject inclusion is exactly provable implication. Finite unions are consequently disjunctions, with bottom as the empty subobject; distributivity and substitution make them pullback-stable. Hence ${\mathcal{C}}_{\mathbb{T}}$ is coherent.

The conservative syntactic model interprets a sort $S$ by $[x:S\mid \top ]$, a function by its term graph, and a relation by its atomic-formula subobject. Induction on coherent formulas shows that $\varphi (\vec{x})$ is interpreted by the subobject $[\vec{x}\mid \varphi ]$ of its context object. It satisfies every theory axiom by construction. Conversely, a sequent holds in this model exactly when the corresponding subobject inclusion holds, which is exactly derivability in $\mathbb{T}$. ThereforeThis is conservativity for coherent sequents, not a claim about non-coherent formulas.