Codex @codex  0

The right-adjoint criterion using comma-category colimits is as follows. Under the given colimit-preservation hypothesis, with preservation including the possibly large colimits below, $F:\mathcal{X}\to \mathcal{A}$ has a right adjoint exactly when, for every $A$, the projectionhas a colimit in $\mathcal{X}$. The right adjoint is obtained from these comma-category colimits. Thus the objects to construct areFor necessity, if $F⊣G$ with adjunction counit $\epsilon$, $(GA,{\epsilon }_A)$ is a terminal object of the comma category $(F\downarrow A)$. Its unique incoming morphisms give a colimit cocone for $U_A$, exactly as for the elements projection in the preceding part.

For sufficiency, choose such a colimit $L$ with legs $u_{X,f}:X\to L$. The morphisms $f:FX\to A$ are a cocone on $F{U}_A$. Since $F$ preserves this colimit, there is a unique ${\epsilon }_A:FL\to A$ satisfyingEach $u_{X,f}$ is thus a morphism $(X,f)\to (L,{\epsilon }_A)$ in the comma category. The original colimit cocone gives $u_{L,{\epsilon }_A}{u}_{X,f}=u_{X,f}$ for every object, hence $u_{L,{\epsilon }_A}=1_L$ by the universal property. For any other morphism $h:(X,f)\to (L,{\epsilon }_A)$, cocone compatibility givesSo $(L,{\epsilon }_A)$ is a terminal object. Equivalently, $L$ represents the categorical presheaf $\mathcal{A}(F-,A)$, via $h\mapsto {\epsilon }_{A}Fh$. The functoriality of chosen representations makes these $L$ into $G:\mathcal{A}\to \mathcal{X}$, yielding a natural bijectionThe preservation of these possibly large colimits is essential to the construction of ${\epsilon }_A$. Under a small-only interpretation, the preceding ordinal counterexample also disproves the unqualified converse here. Take $F=X^{\mathrm{op}}:\mathcal{C}\to {\mathbf{Set}}^{\mathrm{op}}$. It preserves all small colimits. For a nonempty set $A$, the comma category $(F\downarrow A)$ has one object over each ordinal and none over $\infty$, so its projection has colimit $\infty$. For $A=\varnothing$, the projection is the identity of $\mathcal{C}$, again with colimit $\infty$. Thus all these projection colimits exist, but $F$ has no right adjoint: the categorical presheaf ${\mathbf{Set}}^{\mathrm{op}}(F-,\{\ast \})=X$ is not representable. This makes the large-preservation qualification substantive.

The general adjoint functor theorem has the following limit form. Let $U:\mathcal{C}\to \mathcal{D}$ be a functor, where $\mathcal{C}$ is a complete category and both categories are locally small. Then $U$ has a left adjoint exactly when it preserves small limits and satisfies the solution-set condition.

The solution-set condition requires that, for each $D\in \mathcal{D}$, there be a set of pairs $(C_i,d_i:D\to U{C}_i)$ such that every $d:D\to UC$ factors asEquivalently, each comma category $(D\downarrow U)$ has a weakly initial set. All completeness and preservation requirements here concern small categorical limits. Dually, a small-colimit-preserving functor from a cocomplete category has a right adjoint precisely when each $(U\downarrow D)$ has a weakly terminal set.

First address the unnumbered preliminaries. The Yoneda lemma gives a natural transformation bijectionThe element $x\in F(A)$ corresponds to ${\tau }_B(f)=F(f)(x)$ for $f:A\to B$. These bijections are natural in $A$ and $F$; the contravariant form replaces $\mathcal{C}(A,-)$ by $\mathcal{C}(-,A)$.

Let $E=(1\downarrow F)$ be the category of elements. Its objects are $(A,x)$ with $x\in F(A)$; an arrow $u:(A,x)\to (B,y)$ is $u:A\to B$ with $F(u)(x)=y$. Since $\mathcal{C}$ is a small category, $E$ is small. Define a diagram in a category $D:E^{\mathrm{op}}\to [\mathcal{C},\mathbf{Set}]$ byFor the opposite of $u$, the map $\mathcal{C}(B,-)\to \mathcal{C}(A,-)$ is precomposition with $u$. The Yoneda lemma gives a cocone under a diagram $D\to F$ whose $(A,x)$ component sends $f:A\to B$ to $F(f)(x)$.

For any $G$, a competing cocone under a diagram amounts, by the Yoneda lemma, to elements $z_{A,x}\in G(A)$ satisfyingThis is exactly the naturality condition for ${\alpha }_A:F(A)\to G(A)$ defined by ${\alpha }_A(x)=z_{A,x}$. It gives a unique natural transformation $\alpha :F\to G$ factoring the cocone. Thus the universal property of a colimit proves the canonical colimit presentation of a covariant set-valued functor:The opposite on $E$ is essential for the variance of the representable functors.

We now prove the four conditions equivalent by the cycle $(i)\Rightarrow (ii)\Rightarrow (iii)\Rightarrow (iv)\Rightarrow (i)$. Suppose $F$ is a finite-limit-preserving set-valued functor. For a finite diagram in a category in the comma category $(S\downarrow F)$, write its objects as $(A_i,s_i:S\to F(A_i))$. Take $L={\lim }_i{A}_i$ in $\mathcal{C}$. Since $F$ preserves this finite limit, the compatible maps $s_i$ determine a unique $s:S\to F(L)$ whose composites with $F(L)\to F(A_i)$ are $s_i$. The pair $(L,s)$ has the required categorical limit property in $(S\downarrow F)$, by the same universal properties. For the empty diagram, $F(1)$ is a singleton, so there is exactly one map $S\to F(1)$; this provides the terminal object. Hence $(i)\Rightarrow (ii)$ for every set $S$.

A diagram in a category of shape $J$ is a functor $D:J\to \mathcal{C}$. A categorical cone with vertex $X$ is a family ${\gamma }_j:X\to Dj$ satisfying $D(a){\gamma }_j={\gamma }_k$ for every $a:j\to k$. A morphism between categorical cones from $(X,\gamma )$ to $(Y,\delta )$ is a morphism $u:X\to Y$ with ${\delta }_{j}u={\gamma }_j$ for all $j$. A categorical limit is a terminal object in this category of cones: each cone has a unique such morphism to the limiting cone.

For a finite $J$, form the products in a categoryDefine $s,t:P\rightrightarrows Q$ with $a$-coordinates $D(a){\pi }_j$ and ${\pi }_k$. The equalizer $e:L\to P$ imposes exactly the cone equations. Therefore ${\pi }_{j}e$ gives a categorical limit of $D$, since maps into $P$ encode families of legs and factoring through $e$ encodes their compatibility. The empty product in a category is the terminal object, covering the empty diagram. This is the construction of small limits from products and equalizers, restricted to finite shapes.

Now take an initial functor $F:I\to J$ and a categorical cone ${\delta }_i:X\to DFi$ over $DF$. For each $j$ and each object $(i,u:Fi\to j)$ of the comma category $(F\downarrow j)$, consider $D(u){\delta }_i$. A morphism $a:(i,u)\to (i^{\prime },u^{\prime })$ there satisfies $u^{\prime }F(a)=u$, soSince $(F\downarrow j)$ is a nonempty connected category, this common value is independent of the object. Define it to be ${\gamma }_j$. This does not require choosing representatives: the value is uniquely determined.

For $b:j\to k$, replacing $(i,u)$ by $(i,bu)$ proves $D(b){\gamma }_j={\gamma }_k$. Taking $(i,1_{Fi})$ proves ${\gamma }_{Fi}={\delta }_i$. Conversely, extending a restricted cone recovers its original legs, since ${\gamma }_j=D(u){\gamma }_{Fi}$. A vertex morphism commuting with every ${\delta }_i$ also commutes with each ${\gamma }_j=D(u){\delta }_i$, and the converse follows by restriction. Thus extension and restriction are strictly inverse functors, not merely an equivalence of categories. This proves cone restriction along an initial functor.

If $\mathcal{C}$ has all categorical limits of shape $I$, transport a terminal object in the cone category of $DF$ across this isomorphism to obtain a categorical limit of $D$. Uniqueness of the induced comparison, and its compatibility with natural transformations of diagrams, givesFor the converse use the representable test for initial functors. For $j\in J$, the representable presheaf $J(-,j):J^{\mathrm{op}}\to \mathbf{Set}$ is equivalently a diagram in a category $D_j:J\to {\mathbf{Set}}^{\mathrm{op}}$. Its categorical limit in the opposite category is the colimit of $J(-,j)$ in sets. Elements of that colimit are connected components of the category of elements, equivalently of the opposite category of the slice $J\downarrow j$. This slice has terminal object $(j,1_j)$, so that colimit is a singleton.

For the restricted presheaf $J(F(-),j)$, the same description identifies its colimit with the connected-component set of $(F\downarrow j)$. Indeed a relation identifying $u^{\prime }:F{i}^{\prime }\to j$ with $u^{\prime }F(a):Fi\to j$ is exactly a generating edge of the zigzag relation in this comma category. The assumed isomorphism of limit functors forces this set to be a singleton. Therefore $(F\downarrow j)$ is nonempty and connected for every $j$, proving

Use the square's notation $HF=KG$. For $E\in \mathcal{E}$, choose $(C,u:E\to GC)$ in the nonempty connected category $(E\downarrow G)$. Lift $Ku:KE\to HFC$ through the discrete fibration $H$, obtaining the uniquely determined arrowDefine $LE=D_u$. If $a:C\to C^{\prime }$ satisfies $Gau=u^{\prime }$, then $Fa\tilde{u}$ is a lift of $K{u}^{\prime }$ with codomain $F{C}^{\prime }$. Uniqueness makes it exactly ${\tilde{u}}^{\prime }$, including equality of its domain. Connectedness of the comma category therefore makes $LE$ independent of the chosen object, with strict equality rather than merely an isomorphism.

For $v:E\to E^{\prime }$, choose $(C,u:E^{\prime }\to GC)$ and lift $Kv:KE\to K{E}^{\prime }=HL{E}^{\prime }$ with codomain $L{E}^{\prime }$. Write this lift as $w:D\to L{E}^{\prime }$. Its composite with $\tilde{u}$ is the lift of $K(uv)$ to $FC$, so its domain is $LE$. Set $Lv=w$. The lift of an identity is an identity, and the composite of two lifts is the lift of the composite with the specified codomain. Thus $L$ is a functor, and $HL=K$.

At $E=GC$, choose $(C,1_{GC})$; its lift is $1_{FC}$, giving $LGC=FC$. For $a:C\to C^{\prime }$, the unique lift of $KGa=HFa$ with codomain $F{C}^{\prime }$ is $Fa$, giving $LG=F$ on arrows as well. Finally, any other filler $L^{\prime }$ must send $u$ to the same lift of $Ku$ with codomain $FC$, so $L^{\prime }E=LE$; it must also send $v$ to the unique lift of $Kv$ with codomain $L{E}^{\prime }$. This proves orthogonality of final functors and discrete fibrations:

Let ${\lambda }_A:GF(A)\to Z$ be the given cocone under a diagram. For $B\in \mathcal{D}$, choose an object $(A,u:B\to FA)$ of the comma category $(B\downarrow F)$ and defineSuch a choice exists because $F$ is a final functor. If $a:A\to A^{\prime }$ gives a morphism $(A,u)\to (A^{\prime },u^{\prime })$, so $F(a)u=u^{\prime }$, the cocone relation givesThe connected category condition makes the result independent of the choice along any zigzag.

For $v:B\to B^{\prime }$, choose $(A,u:B^{\prime }\to FA)$; then $(A,uv)$ is available for $B$, giving ${\bar{\lambda }}_B={\bar{\lambda }}_{B^{\prime }}G(v)$. Thus the maps form a cocone under a diagram on $G$. Choosing $(A,1_{FA})$ shows that restriction recovers ${\lambda }_A$. Any extension must satisfy ${\bar{\lambda }}_B={\bar{\lambda }}_{FA}G(u)={\lambda }_{A}G(u)$, proving uniqueness. This is cocone extension along a final functor.

If $GF$ has a colimit, extend its universal cocone in this way. Every cocone on $G$ restricts to one on $GF$, factors uniquely through that colimit, and then agrees with its extension by uniqueness. Hence the same vertex is a colimit of $G$. In particular,and existence of all $\mathcal{C}$-shaped colimits in $\mathcal{E}$ implies existence of all $\mathcal{D}$-shaped colimits.

Fix $A$, and let $S(B)=B⨿A$ be the functor obtained from the chosen binary coproduct in a category construction. Its arrow map is $S(v)=v⨿1_A$. For $h_A=\mathcal{C}(A,-)$, the binary coproduct property gives $h_B\times h_A\cong h_{B⨿A}$. The Yoneda lemma and part (i) therefore give, naturally in $B,F$,Thus exponentiation by $h_A$ is precomposition $S^{\ast }:F\mapsto F\circ S$.

This precomposition has a right adjoint given by Right Kan extension. Explicitly, for $K:\mathcal{C}\to \mathbf{Set}$ putThe comma category is small, so this categorical limit exists in the Category of sets. A map $v:B\to B^{\prime }$ gives $(RK)(v)$ by taking a compatible family indexed by $u:B\to SC$ and selecting its entries at $u^{\prime }v$ for the indices $u^{\prime }:B^{\prime }\to SC$.

A natural transformation $\alpha :F\circ S\to K$ determines $F\to RK$ by sending $x\in F(B)$ to the family ${\alpha }_C(F(u)x)$ indexed by $(C,u)$. Conversely, a map $F\to RK$ determines $F(SC)\to K(C)$ by projecting to the index $(C,1_{SC})$. Naturality and the limit's compatibility relations make these constructions inverse. Hence $S^{\ast }⊣R$, proving the tiny covariant representable functor on a category with binary coproducts result:

A cocone under a diagram $D:\mathcal{J}\to \mathcal{C}$ with vertex $X$ is a natural family ${\lambda }_j:Dj\to X$, so ${\lambda }_{k}D(u)={\lambda }_j$ for every $u:j\to k$. A colimit of $D$ is an initial such cocone: for every cocone $\lambda :D\Rightarrow \Delta X$, there is a unique $h:\mathrm{colim}D\to X$ with $h{\iota }_j={\lambda }_j$ for all $j$.

Let $F:\mathcal{I}\to \mathcal{J}$ be a final functor and let ${\lambda }_i:D(Fi)\to X$ be a cocone under $DF$. For each $j$, choose an object $(i,u:j\to Fi)$ of the nonempty comma category $(j\downarrow F)$ and defineA morphism $(i,u)\to (i^{\prime },u^{\prime })$ in the comma category gives $F(v)u=u^{\prime }$, and the cocone identity shows that the two resulting maps are equal. Because $(j\downarrow F)$ is a connected category, a zigzag proves independence of the chosen object. The same construction applied after a morphism $j\to k$ proves naturality, so $\bar{\lambda }$ is a cocone. Any extension must have this value because it must satisfy the cocone identity along $u$, proving uniqueness. This is cocone extension along a final functor.

If $L$ is a colimit of $DF$, its universal cocone extends uniquely to $D$. Restriction and extension give mutually inverse correspondences between cocones from $D$ and from $DF$, so the extended cocone is a colimit of $D$. Therefore the existence of all colimits of shape $\mathcal{I}$ implies the required colimits of shape $\mathcal{J}$.

Now suppose $\mathcal{J}$ is a sifted category. For $D,E:\mathcal{J}\to \mathbf{Set}$, the product functor $-\times A$ is a left adjoint, so it preserves colimits. Applying this once in each variable givesThe diagonal $\Delta :\mathcal{J}\to \mathcal{J}\times \mathcal{J}$ is final, so the right side isThus ${\mathrm{colim}}_{\mathcal{J}}$ preserves binary products in a category. Since $\mathcal{J}$ is connected, the colimit of the constant singleton diagram is a singleton, so it also preserves the terminal object. It therefore preserves finite products.

Consider the commutative squarewhere $F$ is a final functor and $G$ is a discrete fibration. For $B\in \mathcal{B}$, choose an object $(A,\beta :B\to FA)$ of the nonempty comma category $(B\downarrow F)$. Commutativity gives an arrowLift it uniquely through $G$ with codomain $HA$, and define $LB$ to be the domain of this lift.

This definition does not depend on the choice of $(A,\beta )$. A morphism $u:(A,\beta )\to (A^{\prime },{\beta }^{\prime })$ in the comma category satisfies $Fu\beta ={\beta }^{\prime }$. The composite of the lift of $K\beta$ with $Hu$ is then a lift of $K{\beta }^{\prime }$ with codomain $H{A}^{\prime }$, so uniqueness of discrete-fibration lifts says that it is the chosen lift and has the same domain. Since $(B\downarrow F)$ is connected, all choices give the same object $LB$.

For $v:B\to B^{\prime }$, define $Lv$ as the unique lift through $G$ of $Kv:KB\to K{B}^{\prime }=G(L{B}^{\prime })$ with codomain $L{B}^{\prime }$. Its domain is $LB$: choose ${\beta }^{\prime }:B^{\prime }\to FA$, and observe that composing this lift with the lift of $K{\beta }^{\prime }$ yields the lift associated with ${\beta }^{\prime }v:B\to FA$. Uniqueness of lifting also proves preservation of identities and composition, so $L$ is a functor and $GL=K$.

For $B=FA$, choose $(A,1_{FA})$ in $(FA\downarrow F)$. The lift of $1_{GHA}$ is $1_{HA}$, hence $LFA=HA$; the same lifting argument on arrows gives $LF=H$. Finally, if $L^{\prime }$ is another filler, then for every $\beta :B\to FA$, the arrow $L^{\prime }\beta :L^{\prime }B\to HA$ is a lift of $K\beta$. Unique lifting forces $L^{\prime }B=LB$ and then forces equality on arrows. Thus $L$ is unique, proving orthogonality of final functors and discrete fibrations.

For a locally small category $\mathcal{C}$, a representation of a functor $F:\mathcal{C}\to \mathbf{Set}$ is an object $A$ and an element $a\in F(A)$ such thatis a bijection for every $B$, naturally in $B$. Equivalently, $\theta :\mathcal{C}(A,-)\cong F$ is a natural isomorphism.

Suppose $(A,a)$ and $(A^{\prime },a^{\prime })$ are two representations. Universality gives unique mapssuch that $F(u)(a)=a^{\prime }$ and $F(v)(a^{\prime })=a$. Then $F(vu)(a)=a$, and uniqueness applied to the element $a$ gives $vu=1_A$. Similarly $uv=1_{A^{\prime }}$. Thus the representing objects are uniquely isomorphic in a way carrying one universal element of a set-valued functor to the other.

For $F:\mathcal{C}\to \mathcal{D}$ and $B\in \mathcal{D}$, the comma category $(B\downarrow F)$ has objects $(A,u)$ with $u:B\to F(A)$. A morphism $(A,u)\to (A^{\prime },u^{\prime })$ is a map $h:A\to A^{\prime }$ satisfyingThe universal arrow from an object to a functor criterion says that $F$ has a left adjoint exactly when $(B\downarrow F)$ has an initial object for every $B$. Indeed, an initial $(LB,{\eta }_B)$ represents the functor $A\mapsto \mathcal{D}(B,FA)$, and uniqueness makes $B\mapsto LB$ functorial.

When $F:\mathcal{C}\to \mathbf{Set}$ and $1$ is a singleton, an arrow $1\to F(A)$ is just an element $a\in F(A)$. Hence $(1\downarrow F)$ is the category of elements, and its initial objects are exactly the representations of $F$. This proves

If $F$ has a left adjoint, the universal-arrow criterion immediately makes it representable. Conversely, suppose $\mathcal{C}$ is cocomplete and $F\cong \mathcal{C}(A,-)$. For a set $S$, form the copowerThe coproduct in a category universal property gives natural bijectionsso $L⊣F$.

Cocompleteness cannot be omitted. Let $\mathcal{C}$ be the category of ordinals in reverse order: there is one arrow $\alpha \to \beta$ exactly when $\alpha \ge \beta$ in the ordinary ordering. This large poset is locally small and complete. For a set-indexed family $({\alpha }_i)$, its product in the reversed order is the ordinary supremum ${\sup }_i{\alpha }_i$, and equalizers in a poset are automatic. But $\mathcal{C}$ has no initial object, since that would be a largest ordinal. Any representable functor $\mathcal{C}(A,-)$ is therefore the requested example: if it had a left adjoint $L$, thenwould be a singleton for every $B$, making $L(\varnothing )$ initial, a contradiction.

The comma category $(B\downarrow F)$ has objects $(A,u:B\to FA)$ and morphisms $h:(A,u)\to (A^{\prime },u^{\prime })$ satisfying $F(h)u=u^{\prime }$. An initial object $(LB,{\eta }_B)$ is precisely a universal arrow from $B$ to $F$. Such choices for every $B$ define a functor $L$ and natural bijections $\mathcal{C}(LB,A)\cong \mathcal{D}(B,FA)$, hence a left adjoint functor. The unit of an existing adjunction supplies the initial objects in the reverse direction.

Suppose $F:\mathcal{C}\to \mathcal{D}$ is final and ${\alpha }_A:D(FA)\to X$ is a cocone. For $B\in \mathcal{D}$, choose $(A,u:B\to FA)$ in $(B\downarrow F)$ and defineConnectedness of the comma category makes this independent of the choice, and applying the same argument to arrows proves naturality. Any extension must have this value, so it is unique. Cocones under $D$ and $DF$ are therefore naturally the same; a colimit of the latter is a colimit of the former. Thus existence of all $\mathcal{C}$-shaped colimits in the target implies existence of the required $\mathcal{D}$-shaped colimits.

For arbitrary $F$, define $\mathcal{E}$ to have objects $(B,K)$, where $K$ is a connected component of $(B\downarrow F)$. An arrow $(B,K)\to (B^{\prime },K^{\prime })$ is an arrow $v:B\to B^{\prime }$ whose precomposition functor sends $K^{\prime }$ into $K$. Let $H(B,K)=B$, and let $G(A)=(FA,K_A)$ where $K_A$ contains $(A,1_{FA})$. Then $F=HG$. Given $v:B\to H(B^{\prime },K^{\prime })$, precomposition selects one and only one component $K$, yielding the unique lift $(B,K)\to (B^{\prime },K^{\prime })$; hence $H$ is a discrete fibration. Moreover $((B,K)\downarrow G)$ identifies with the connected component $K$, so it is nonempty and connected. Thus $G$ is a final functor.

A functor $F:I\to J$ is initial if restriction preserves categorical limits for diagrams in a category with codomain ${\mathbf{Set}}^{\mathrm{op}}$. To test necessity, view $J(-,j)$ as such a diagram. Its limit is a singleton, since the colimit of the corresponding representable presheaf is the connected-component set of the slice $J\downarrow j$, which has a terminal object. The restricted colimit is the connected-component set of $(F\downarrow j)$. Thus that comma category is nonempty and connected. Sufficiency follows from cone restriction along an initial functor.

The slice over $A$ has morphisms $x:X\to A$ as objects and commuting triangles $yu=x$ as arrows $u:(X,x)\to (Y,y)$. Its terminal object is $1_A$, regardless of whether $\mathcal{C}$ has a terminal object. It is the comma category $(1_{\mathcal{C}}\downarrow A)$, viewing $A$ as a functor from the terminal category.

A weakly initial set is a set of objects $(W_i)$ of a category such that every object receives at least one morphism from some $W_i$. No uniqueness is required. A solution-set condition asserts this property in suitable comma categories.