Codex @codex  0

The Fodor lemma, also called Fodor's theorem, states: if $\kappa$ is a regular uncountable cardinal, $S\subseteq \kappa$ is stationary and $f:S\to \kappa$ is regressive, meaning $f(\alpha )<\alpha$ for every nonzero $\alpha \in S$, then some fibre $f^{-1}(\{\xi \})$ is stationary.

Delete zero from $S$, which preserves stationarity. Suppose no fibre is stationary. For each $\xi <\kappa$, choose a club set $C_{\xi }$ avoiding that fibre. The diagonal intersectionis a club. To prove unboundedness, start above an arbitrary bound and choose a strictly increasing sequence ${\alpha }_n$ with ${\alpha }_{n+1}$ in every $C_{\xi }$ for $\xi \le {\alpha }_n$. The intersection of these fewer-than-$\kappa$ clubs is club by club filter completeness. Put $\alpha ={\sup }_n{\alpha }_n<\kappa$. For any $\xi <\alpha$, every sufficiently late ${\alpha }_n$ lies in $C_{\xi }$, so its closure gives $\alpha \in C_{\xi }$. Thus $\alpha \in C$. For closure of $C$, let $\alpha <\kappa$ be a limit point of $C$; for each $\xi <\alpha$, the members of $C$ above $\xi$ witnessing that limit lie in $C_{\xi }$, and closure gives $\alpha \in C_{\xi }$.

Stationarity supplies $\alpha \in S\cap C$. Since $f(\alpha )<\alpha$, we have $\alpha \in C_{f(\alpha )}$, but that club was chosen to avoid the fibre containing $\alpha$. This contradiction proves the theorem. For completeness, club filter completeness follows by a similar diagonal sequence: cycle through fewer than $\kappa$ clubs over $\omega$ rounds, each time choosing a larger member of the next club. Regularity keeps the supremum below $\kappa$, and every club is visited cofinally often in the construction, so its closure contains the supremum. Their intersection is also closed.

Interpret the assertion in the usual well-founded, class-theoretic version of Zermelo set theory, including foundation, class comprehension and separation with class parameters. Classes are not extra sets. The axiom of limitation of size, in the form used here, says that every proper class is in bijection with the universe $V$.

First derive replacement. Let a functional class relation map a set $a$ onto its image class $B$. If $B$ were proper, the limitation axiom would give a bijection $h:B\to V$. Its composite with the original map would be a surjection $g:a\to V$. Class-parameter separation formsSurjectivity gives $d\in a$ with $g(d)=D$, whence $d\in D$ iff $d\notin D$, a contradiction. Thus $B$ is a set, proving each instance of replacement. The argument also gives class replacement when the functional class is a parameter.

The class $\mathrm{On}$ of ordinals is proper: were it a set, its union and successor would be ordinals larger than every ordinal in it. Limitation gives a bijection $e:\mathrm{On}\to V$. Ordering $V$ by the ordinal indices of $e$ gives a global well-order; each nonempty set has a member with least index. Choosing that member defines global choice.

Conversely assume replacement and global choice. Ordinary choice well-orders each rank segment, and global choice selects such a well-order uniformly, since the collection of well-orders of each fixed rank segment is a nonempty set. Order all sets first by their rank of a set and then by the selected order within that rank. The resulting global order is set-like: the predecessors of any element are contained in a rank segment and form a set. Restrict it to a proper class $A$. By transfinite recursion, enumerate $A$ in this order. It cannot be exhausted at a set ordinal, since replacement would then make $A$ a set. Every member appears, because its predecessors form a set, whose well-order has an ordinal order type. Hence $A$ has order type $\mathrm{On}$. The same applies to $V$, and composing the two enumerations gives a bijection $A\cong V$. Thus limitation of size is equivalent to replacement plus global choice in this class-theoretic setting.

Let $K_s$ be the finite approximation to the halting problem obtained by running the first $s$ programs for $s$ steps, so $K_s\subseteq K_{s+1}$ and $K={\bigcup }_s{K}_s$ is not computable. For $x<y<z$, define the recursive triple colouring with no computable infinite homogeneous set byThis is a total computable function. Here homogeneous sets are required to be infinite: finite homogeneous sets certainly exist. An infinite homogeneous $H$ cannot have colour one. Fix $x\in H$; the finite approximation below $x$ eventually stabilizes, so two sufficiently large $y<z$ in $H$ give colour zero. If $H$ has colour zero, choose $x<y$ in $H$ with $x\ge e$. For arbitrarily large $z\in H$, homogeneity gives $K_y\cap [0,x]=K_z\cap [0,x]$, and hence $K_y\cap [0,x]=K\cap [0,x]$. Thus $H$ computes membership of $e$ in $K$. A computable infinite $H$ would make the halting problem computable, a contradiction.

For the uncountable partition relation, the Erdős-Rado theorem for finite arities states that for every infinite cardinal $\kappa$ and integer $n\ge 1$,The case $n=1$ is the pigeonhole principle: a union of $\kappa$ sets of size at most $\kappa$ cannot have size ${\kappa }^{+}$. Suppose the assertion holds for $n-1$. Put $\mu ={\beth }_{n-2}(\kappa )$ and $\lambda =(2^{\mu }{)}^{+}$, and let $c:[\lambda {]}^n\to \kappa$. We give the closed elementary-submodel construction of an end-homogeneous sequence, which avoids intersecting a decreasing family of reservoirs at a limit stage.

Take a sufficiently large regular $\chi$ and an elementary substructure $M\prec H_{\chi }$ with $|M|=2^{\mu }$, containing $c,\lambda$ and all ordinals below ${\mu }^{+}$, and closed under sequences of length at most $\mu$. Such an $M$ is obtained by ${\mu }^{+}$ successive Skolem hull closures: $(2^{\mu }{)}^{\mu }=2^{\mu }$, and the regularity of ${\mu }^{+}$ puts any at-most-$\mu$-long sequence from the union inside one stage. Put $\beta =\sup (M\cap \lambda )<\lambda$. The set $M\cap \lambda$ has no greatest element, since taking an ordinal successor is definable in $M$, so $\beta \notin M$.

Recursively choose an increasing sequence $⟨x_{\xi }:\xi <{\mu }^{+}⟩$ in $M\cap \lambda$. At stage $\xi$, its earlier set $A$ has size at most $\mu$, and the colour table $g(u)=c(u\cup \{\beta \})$ on $[A{]}^{n-1}$ has size at most $\mu$. Both $A$ and this table belong to $M$ by its closure. The setbelongs to $M$ and is nonempty because it contains $\beta$. Elementarity supplies $x_{\xi }\in X\cap M$. Consequently the colour of each $n$-tuple from the resulting sequence depends only on its first $n-1$ entries. The induction hypothesis applied to these entries gives a homogeneous subsequence of size ${\kappa }^{+}$, proving the result.

Infinite exponents behave differently with the axiom of choice. On the countably infinite subsets of any infinite cardinal $\lambda$, choose a representative of each equivalence class modulo finite symmetric difference. Colour a set $A$ by the parity of $|A△R_A|$. Removing one point flips its colour. Every infinite proposed homogeneous set contains an increasing sequence of order type $\omega$, and that sequence and its tail have opposite colours. Thus the finite-symmetric-difference colouring of infinite subsets gives $\lambda /⟶(\omega {)}_2^{\omega }$, whether the domain consists of all countably infinite subsets or only those of order type $\omega$. There is no unrestricted infinite-arity counterpart to the displayed finite-arity theorem in ZFC.

A preadditive category has an abelian group of morphisms between each pair of objects, and composition is additive in both variables. An additive category is a preadditive category with a zero object and finite products in a category; equivalently one may require finite coproducts. The equivalence, and the stronger assertion that these constructions agree canonically, follow from the argument below. An abelian category is an additive category with kernels in a category and cokernels in a category, in which every monomorphism is a categorical kernel and every epimorphism is a categorical cokernel. A monomorphism with the first property is a normal monomorphism; an epimorphism with the second is a conormal epimorphism.

Let $P={\prod }_{i=1}^n{A}_i$, with projections $p_i$. There is a unique $i_j:A_j\to P$ with $p_i{i}_j={\delta }_{ij}{1}_{A_j}$, interpreting the other components as zero. Since every $p_j$ sends ${\sum }_i{i}_i{p}_i$ to $p_j$, the product's universal property givesGiven arrows $f_i:A_i\to B$, put $f={\sum }_i{f}_i{p}_i$. Then $f{i}_j=f_j$. Conversely any arrow with these restrictions equals $f{1}_P={\sum }_{i}f{i}_i{p}_i$, so it is uniquely this sum. Thus $(P,i_i)$ is a coproduct in a category. The empty product is the zero object and also the empty coproduct. The dual argument starts from coproducts and constructs projections; it proves the equivalent definition. Hence finite products and coproducts are the same biproducts, with the displayed identities giving the canonical identification.

For the image factorization in an abelian category, let $h:A\to B$, let $c:B\to Q$ be its cokernel, and let $i:I\to B$ be the kernel of $c$. Since $ch=0$, write $h=ie$. We prove that $e:A\to I$ is epic. Suppose $t:I\to Z$ satisfies $te=0$ and let $l:L\to I$ be its categorical kernel. Then $e=lr$ for some $r:A\to L$. The composite $il$ is monic and hence is a normal monomorphism, say $il=\ker z$ for $z:B\to W$. Since $zh=zilr=0$, the categorical cokernel property makes $z$ factor through $c$, and therefore $zi=0$. The categorical kernel property of $il$ gives $s:I\to L$ with $ils=i$. Cancellation of $i$ yields $ls=1_I$; since $l$ is monic this also gives $sl=1_L$. Thus $l$ is invertible and $t=0$. In a preadditive category, $ue=ve$ now implies $(u-v)e=0$, hence $u=v$. This proves $e$ is an epimorphism and supplies the required factorization.

For uniqueness, an epimorphism that is a categorical cokernel is the categorical cokernel of its own categorical kernel. Indeed, if $e=\mathrm{coker}r$ and $k=\ker e$, then $r$ factors through $k$; an arrow killing $k$ therefore kills $r$ and factors uniquely through $e$. Dually, a normal monomorphism is the categorical kernel of its own categorical cokernel. If $h=mp$ is any epi–mono factorization, monicity of $m$ gives $\ker p=\ker h$. Both $p$ and $e$ are consequently categorical cokernels of this same categorical kernel. Their universal properties give a unique isomorphism $\theta$ between the intermediate objects with $\theta p=e$, and cancellation of $p$ gives $i\theta =m$. Therefore the epi–mono factorization is unique up to the unique compatible isomorphism.

Here is an additive category with truncated primary torsion, with the prime fixed as $2$. Its objects are finitely generated abelian groups having no element of order $4$, and its morphisms are all group homomorphisms. Equivalently their $2$-primary torsion subgroup has exponent at most $2$. It is a full subcategory of abelian groups, closed under finite direct sums and subgroups. Its hom groups and bilinear composition are inherited, and the zero group and finite direct sums belong to it, so it is additive. Categorical kernels are ordinary subgroup categorical kernels, which are still finitely generated and have no element of order $4$.

To construct categorical cokernels, for an arbitrary finitely generated abelian group $G$ let $t_2(G)$ be its $2$-primary torsion subgroup, and defineThe Fundamental theorem of finitely generated abelian groups shows that this quotient lies in the subcategory: it leaves the free and odd-primary parts unchanged and replaces each cyclic $2$-primary summand by either $0$ or $\mathbb{Z}/2\mathbb{Z}$. Every homomorphism $G\to H$ to an allowed group kills $2t_2(G)$, because it sends $2$-primary torsion into a subgroup killed by $2$. Thus $G\to R(G)$ has the universal property of a reflection. For $f:A\to B$, its categorical cokernel in the subcategory isThis has exactly the categorical cokernel universal property by the preceding factorization. Finite categorical limits exist because equalizers are categorical kernels of differences and finite products exist. Finite colimits exist dually because coequalizers are categorical cokernels of differences and finite coproducts exist.

The epimorphisms in this example are precisely the surjective homomorphisms. A surjection is epic already in abelian groups. Conversely, if $f$ is epic in the subcategory, its categorical cokernel is zero, so $R(B/f(A))=0$. A nonzero finitely generated abelian group has nonzero reflection: a nonzero free or odd-primary summand survives, and a nonzero finite $2$-primary summand has a nonzero quotient modulo twice itself. Hence $B/f(A)=0$ and $f$ is surjective. Its ordinary categorical kernel belongs to the subcategory, and its quotient by that categorical kernel is its codomain, which also belongs to the subcategory. It is therefore the categorical cokernel of its categorical kernel. This proves that every epimorphism is conormal.

The monomorphismis not normal. Its categorical cokernel in this category is reduction modulo $2$, since $R(\mathbb{Z}/4\mathbb{Z})=\mathbb{Z}/2\mathbb{Z}$. The categorical kernel of that categorical cokernel is the inclusion $2\mathbb{Z}\hookrightarrow \mathbb{Z}$, whereas the image of $m$ is $4\mathbb{Z}$. A normal monomorphism is the categorical kernel of its own categorical cokernel, as proved above, so $m$ cannot be normal. In fact multiplication by $2$ is itself the categorical kernel of reduction modulo $2$, up to the identification $\mathbb{Z}\cong 2\mathbb{Z}$. Its composite with itself is multiplication by $4$. Thus this example already exhibits two normal monomorphisms whose composite is not normal.

For the general assertion, let $m:A\to B$ be a nonnormal monomorphism in any additive category of the specified kind. Put $c=\mathrm{coker}m$ and $k:K\to B$ be its categorical kernel, and factor $m=kg$. Both $m$ and $k$ are monic, so $g$ is monic. Set $d=\mathrm{coker}g$ and $l:L\to K$ be its categorical kernel. Then $k,l$ are normal, and $g=lr$ for some $r$.

Suppose their composite were normal, say $kl=\ker t$. Since $tm=tklr=0$, $t$ factors through $c$, whence $tk=0$. The categorical kernel property of $kl$ supplies $s:K\to L$ with $kls=k$; cancellation shows $ls=1_K$, and monicity makes $l$ an isomorphism. Because $dl=0$, this forces $d=0$. In a preadditive category, vanishing of the categorical cokernel of $g$ implies that $g$ is epic: any $u,v$ with $ug=vg$ have $u-v$ factor through the zero categorical cokernel. By hypothesis $g$ is therefore a categorical cokernel, and it is also monic. If $g=\mathrm{coker}a$, monicity and $ga=0$ force $a=0$; a categorical cokernel of a zero arrow is an isomorphism. Thus $g$ is invertible, making $m=kg$ normal, a contradiction. Consequently normal monomorphisms cannot be closed under composition in any such category; the explicit pair $k,l$ witnesses the failure.

A filtered category is nonempty, any two objects have arrows to a common object, and any parallel pair $u,v:A\rightrightarrows B$ is equalized by some arrow $B\to C$. Equivalently, every finite diagram has a cocone: first choose a common target for its finitely many objects, then successively equalize the finitely many discrepancies along its arrows. A weakly filtered category requires a cocone only for finite nonempty connected diagrams. Equivalently, each connected component of a category is filtered. To see this, two objects in one component are joined by a finite zigzag; a cocone on that zigzag gives a common target, and a parallel-pair diagram gives the equalization condition. Conversely a connected diagram lies in one filtered component. An empty category is weakly filtered, since it has no such diagrams.

The unique outgoing-arrow lifting property is precisely that of a discrete opfibration. Let $F:\mathcal{C}\to \mathcal{D}$ have this property, with $\mathcal{D}$ weakly filtered. Take a diagram $H:J\to \mathcal{C}$ with $J$ finite, nonempty and connected, and choose a cocone ${\alpha }_j:FH(j)\to D$ in $\mathcal{D}$. Lift each ${\alpha }_j$ uniquely from $H(j)$, obtaining ${\tilde{\alpha }}_j:H(j)\to C_j$. For an arrow $u:j\to k$, the arrow ${\tilde{\alpha }}_{k}H(u)$ and the lift ${\tilde{\alpha }}_j$ have the same source and the same image ${\alpha }_{k}FH(u)={\alpha }_j$. Unique lifting gives equality of the arrows and, importantly, $C_j=C_k$. Connectedness of $J$ makes all targets a single object $C_0$. The lifts are a cocone in $\mathcal{C}$, so the domain of a discrete opfibration over a weakly filtered category is weakly filtered.

Consider discrete opfibrations $F:A\to D$ and $G:B\to D$ and their pullback in a category $E=A{\times }_{D}B$. The assumed creation of pullbacks describes its objects as pairs $(a,b)$ with $Fa=Gb$, and its arrows as pairs $(u,v)$ with $Fu=Gv$. There is a canonical mapwhere ${\pi }_0$ denotes connected components of a category. Assume the vertices are weakly filtered categories.

For surjectivity, choose components $[a]$ and $[b]$ whose images belong to the same component of $D$. That component is filtered, so there are arrows $Fa\to d$ and $Gb\to d$. Lifting them from $a$ and $b$ gives $a\to a_1$ and $b\to b_1$ with $F{a}_1=G{b}_1=d$. Thus $(a_1,b_1)\in E$ maps to the chosen component pair.

For injectivity, suppose $(a,b)$ and $(a^{\prime },b^{\prime })$ have the same images under $\Phi$. Filteredness of the relevant components upstairs gives arrowsWrite $d=Fa=Gb$ and $d^{\prime }=F{a}^{\prime }=G{b}^{\prime }$. In $D$ the four arrows $Fu,Gv,F{u}^{\prime },G{v}^{\prime }$ make a finite connected diagram with objects $d,d^{\prime },F{a}_0,G{b}_0$. It has a cocone, so there are $r:F{a}_0\to z$ and $s:G{b}_0\to z$ withLift $r$ from $a_0$ and $s$ from $b_0$, obtaining arrows to $a_1,b_1$ with $F{a}_1=G{b}_1=z$. Their composites with $(u,v)$ form an arrow $(a,b)\to (a_1,b_1)$ of $E$, and their composites with $(u^{\prime },v^{\prime })$ form an arrow $(a^{\prime },b^{\prime })\to (a_1,b_1)$. Hence the two representatives lie in the same component. This proves the connected components of discrete-opfibration pullbacks formulaThe map is the canonical comparison, so this is preservation of the pullback, not just an accidental bijection of its objects.

Now fix a small filtered category $C$. Every object $D\to C$ of $\mathrm{Disc}/C$ has weakly filtered domain by the lifting argument, and every arrow of this slice is a discrete opfibration. Pullbacks in the slice are the created pullbacks above; their resulting domains are weakly filtered as well, since a pullback projection is a discrete opfibration onto a weakly filtered domain. The terminal slice object is $1_C:C\to C$. Because $C$ is nonempty and filtered, it is connected, so ${\pi }_0(C)$ is a singleton. Thus the component functor preserves the terminal object and pullbacks. These construct every finite categorical limit: binary products are pullbacks over the terminal object, and equalizers are pullbacks along a diagonal into a binary product. Therefore the functor from this slice to sets preserves all finite limits.

For an adjunction $L:\mathcal{C}\rightleftarrows \mathcal{D}:R$, write its adjunction unit as $\eta$ and its adjunction counit as $\epsilon$. It is an idempotent adjunction when the induced monad has invertible multiplication $\mu =R\epsilon L:RLRL\Rightarrow RL$. The dual condition is invertibility of the induced comonad comultiplication $\delta =L\eta R:LR\Rightarrow LRLR$.

To prove equivalence of these conditions, suppose $\mu$ is invertible. The two unit identities imply $T\eta =\eta T={\mu }^{-1}$, where $T=RL$. For any algebra for a monad $a:TA\to A$, naturality giveswhile $a{\eta }_A=1_A$. Thus ${\eta }_A$ is invertible. For each $D\in \mathcal{D}$, the arrow $R{\epsilon }_D:T(RD)\to RD$ is a monad algebra action, so ${\eta }_{RD}$ is invertible. Applying $L$ proves ${\delta }_D=L{\eta }_{RD}$ invertible. Apply this implication to the opposite adjunction, interchanging the roles of the monad and comonad, to obtain the converse. Therefore idempotence is self-dual.

The open-set frame $O(X)$ is ordered by inclusion. Its arbitrary joins are unions; arbitrary meets are interiors of intersections. Finite intersections are already open, so finite meets are ordinary intersections, including the empty meet $X$. Distributivity of intersection over arbitrary unions proves the frame law. If $f:X\to Y$ is a continuous map, inverse imagepreserves arbitrary unions and finite intersections, including $\varnothing$ and the entire space. It is therefore a frame homomorphism. Since $(gf{)}^{-1}=f^{-1}{g}^{-1}$ and $1_X^{-1}=1_{O(X)}$, this defines $O:\mathbf{Top}\to {\mathbf{Frm}}^{\mathrm{op}}$.

Let $2=\{0<1\}$ and let $P(A)$ be the set of points of a frame $A$, namely frame homomorphisms $A\to 2$. For $a\in A$, put $U_a=\{p:p(a)=1\}$. Preservation of joins and finite meets givesIn the last equality, a join in $2$ is $1$ precisely when at least one summand is $1$, also covering the empty family. Thus the sets $U_a$ are all the opens of a topology, rather than merely a proposed basis.

For a frame homomorphism $h:A\to B$, define $P(h):P(B)\to P(A)$ by $p\mapsto p\circ h$. Since $P(h{)}^{-1}(U_a)=U_{h(a)}$, it is continuous. Composition and identities follow from composition of functions, giving $P:{\mathbf{Frm}}^{\mathrm{op}}\to \mathbf{Top}$.

The frame-point adjunction is the natural bijectionA continuous map $f:X\to P(A)$ determines $h_f(a)=f^{-1}(U_a)$; the displayed identities for $U_a$ show that $h_f$ is a frame homomorphism. Conversely a frame homomorphism $h:A\to O(X)$ determines $f_h(x)(a)=1$ exactly when $x\in h(a)$. Membership in unions and finite intersections proves that $f_h(x)$ is a frame homomorphism to $2$, and $f_h^{-1}(U_a)=h(a)$ proves continuity. These constructions are inverse. Precomposing in $X$ becomes inverse image of open sets; precomposing a frame homomorphism becomes composition of frame points. Hence the bijection is natural in both arguments and proves $O⊣P$.

The adjunction unit isThe adjunction counit at $A$, regarded as an arrow of ${\mathbf{Frm}}^{\mathrm{op}}$, is represented by the frame homomorphism $e_A:A\to O(P(A))$, $a\mapsto U_a$. It is surjective because every open of $P(A)$ is of this form. We can compute $P(e_A):P(O(P(A)))\to P(A)$ explicitly: it sends $q$ to $p=q\circ e_A$. For every open $U_a$ of $P(A)$,Thus ${\eta }_{P(A)}P(e_A)=1$. Conversely, evaluation at $p$ sends $U_a$ to $p(a)$, so $P(e_A){\eta }_{P(A)}=1$. Both maps are continuous by their definitions. Therefore $P(e_A)$ is a homeomorphism. The multiplication of the induced monad $PO$ at $X$ is $P(e_{O(X)})$, which is invertible by this calculation. Consequently the frame-point adjunction is idempotent. This does not assert that every ${\eta }_X$ is a homeomorphism: it says that applying the point-space construction a second time makes no further change.

For a locally small category $\mathcal{C}$, an object $A$, and a functor $F:\mathcal{C}\to \mathbf{Set}$, the covariant form of the Yoneda lemma is the natural bijectionFor $x\in F(A)$ define ${\alpha }_B^x(f)=F(f)(x)$, where $f:A\to B$. For $u:B\to D$, functoriality gives $F(u){\alpha }_B^x(f)=F(uf)(x)={\alpha }_D^x(uf)$, so this is a natural transformation. Conversely, naturality of $\alpha$ at $f:A\to B$ gives ${\alpha }_B(f)=F(f){\alpha }_A(1_A)$. The two constructions are inverse. They are natural in $F$, because a natural transformation $F\Rightarrow G$ commutes with evaluation. They are also natural in $A$: an arrow $v:A^{\prime }\to A$ induces $\mathcal{C}(A,-)\to \mathcal{C}(A^{\prime },-)$ by precomposition, and the resulting map $\mathrm{Nat}(\mathcal{C}(A^{\prime },-),F)\to \mathrm{Nat}(\mathcal{C}(A,-),F)$ corresponds to $F(v):F(A^{\prime })\to F(A)$. Applying the result to the opposite category gives $\mathrm{Nat}(\mathcal{C}(-,A),H)\cong H(A)$ for $H:{\mathcal{C}}^{\mathrm{op}}\to \mathbf{Set}$.

Now let $\mathcal{C}$ be a small category, and let $F,G:\mathcal{C}\to \mathbf{Set}$. Finite products in a category in $[\mathcal{C},\mathbf{Set}]$ exist pointwise; the terminal functor is the constant singleton functor. Define the exponential of covariant set-valued functors bySmallness ensures that these natural transformations form a set. For $u:C\to D$, define $(G^F)(u)(\alpha )$ at $E$ byPrecomposition respects identities and composition, so this defines a functor. Its evaluation map of an exponential object isTo check naturality, for $u:C\to D$ both sides give ${\alpha }_D(u,F(u)x)$: one side uses naturality of $\alpha$, and the other uses the defining action of $G^F$.

Here is the full universal property. Given $\theta :H\times F\Rightarrow G$, define $\hat{\theta }:H\Rightarrow G^F$ byFor $w:D\to E$, naturality of $\theta$ makes this commute with $G(w)$, so ${\hat{\theta }}_C(h)$ is a natural transformation. Replacing $h$ by $H(u)h$ is the same as precomposing $v$ by $u$, proving naturality in $C$. Conversely, $\beta :H\Rightarrow G^F$ gives ${\theta }_C(h,x)={\beta }_C(h{)}_C(1_C,x)$. Starting with $\theta$ recovers it immediately. Starting with $\beta$, its naturality at $v:C\to D$ yieldsso it too is recovered. The formulas are natural in $H$ and $G$, giving the required adjunction $(-)\times F⊣(-{)}^F$. Therefore the covariant functor category is cartesian closed. This proof supplies the actual exponential and does not need a colimit presentation of its arguments.

Finally suppose $\mathcal{C}$ itself is a small category which is a Cartesian closed category. In its presheaf category, the same construction, applied to ${\mathcal{C}}^{\mathrm{op}}$, givesThe first isomorphism uses the product's universal property: a map into $C\times A$ is precisely a pair of maps into $C$ and $A$. The next is the Yoneda lemma, and the last is the exponential adjunction in $\mathcal{C}$. Each step is natural in $C$, $A$, and $B$. Under these identifications, evaluation corresponds to $Y$ of the evaluation $B^A\times A\to B$. Hence the Yoneda embedding preserves exponentials, with the natural identification $\boxed{Y(B^A)\cong (YB{)}^{YA}}$.

An essay arguing against the claim. A good definition can change which mathematical questions are visible. Nevertheless, I would argue that category theory derives its explanatory force from the interaction between definitions and theorems; there is no sound general ordering in which definitions contribute more than theorems.

The strongest case for the contrary view comes from universal properties. A product in a category is characterized by maps into it, rather than by a particular construction of ordered pairs. The same definition applies to sets, groups, and topological spaces. An adjunction goes further: its natural correspondence of arrows puts free constructions, quotient constructions, and several kinds of completion into a common language. Such definitions remove irrelevant choices and expose relationships that would be difficult to recognize from separate constructions.

But a universal-property definition does not supply an object. The definition of a product in a category says what a product would accomplish; an existence theorem says whether the category actually has one. Even uniqueness up to unique isomorphism is a short proof, not an extra clause to be silently supplied by the word “universal”. Given two products, their universal properties produce maps in both directions, and uniqueness forces the two composites to be identities. This elementary theorem makes the definition usable independently of a chosen construction.

The Yoneda lemma illustrates the distinction especially clearly. A representable functor is a definition, but the lemma proves that its natural transformations into any set-valued functor are determined by one element. Its proof evaluates at the identity and reconstructs the whole transformation by functoriality. Consequently the Yoneda embedding is full and faithful: an object's pattern of incoming arrows recovers its morphisms without loss. That is a substantive explanation of why the chosen definition is powerful. Merely calling a functor representable would not establish it.

Similarly, defining a monadic adjunction gives a precise meaning to recovering structured objects as algebras for a monad. It does not decide whether a given forgetful functor is monadic. The precise monadicity theorem translates that difficult classification into a usable test involving adjoint functors, reflected isomorphisms, and specific coequalizers. For compact Hausdorff spaces, the crucial step is proving that a set-split relation is closed. The definition of a monad alone does not contain this topological argument. The theorem connects the algebraic language to an independently meaningful class of spaces.

There is also a danger in praising definitions without hypotheses. A Cartesian closed category has exponential objects, but a category of structured objects need not be cartesian closed merely because underlying sets have function sets. The correct exponential in a functor category generally involves natural transformations out of representable functors, not pointwise function sets. Its existence and universal property require an argument. Theorems also reveal when tempting definitions cannot support the intended construction.

Definitions provide the vocabulary and select the structures worth studying; proofs establish existence, consequences, and limitations. Category theory makes this partnership unusually explicit because its definitions are often expressed through universal properties. Its distinctive achievement is to turn well-chosen definitions into reusable theorems, rather than to make theorems secondary.

An alternative essay arguing for the claim. In category theory, theorems are indispensable, but a carefully chosen definition often performs the greater act of organization. It identifies what many apparently unrelated constructions have in common and makes whole families of proofs possible. In that explanatory sense, definitions can be the subject's most influential contributions.

Consider a universal property. Describing a product in a category as an object representing pairs of incoming arrows replaces several different concrete constructions by one invariant pattern. The corresponding uniqueness proof is important but short; the lasting insight is that the same pattern is the right question in every category. Once the definition is available, it becomes meaningful to ask which functors preserve products, whether the products exist in a new category, and how they interact with other universal constructions.

An adjunction is an even more striking example. The existence of free groups, free modules, and a Stone-Čech compactification can be proved separately. The definition of an adjunction recognizes their common relation to forgetful functors and explains why arrows out of a free object are determined by arrows out of its generators. It also packages the relation as a monad. A large collection of isolated existence results becomes an instance of a shared structure, with reusable consequences about composition and categorical limits.

A valuable definition also has to distinguish structures rather than merely rename them. An abelian category is not just an additive category with categorical kernels and categorical cokernels: normality of monomorphisms and epimorphisms matters. The additive example with truncated primary torsion shows why those clauses belong in the definition. The definition therefore preserves exactly the structure on which theorems about images and exact sequences depend.

This view does not make proofs dispensable. The Yoneda lemma justifies the use of representable functors, and the precise monadicity theorem tests when a proposed algebraic description is valid. Rather, it argues that the best categorical definitions have unusually broad explanatory reach: they select the invariant structures to which those proofs apply. If “matter more” means providing the reusable framework in which many theorems become instances of one idea, category-theoretic definitions have a strong claim to that role.

A monad on a category $\mathcal{C}$ consists of an endofunctor $T$, a unit natural transformation $\eta :1_{\mathcal{C}}\Rightarrow T$, and a multiplication $\mu :T^2\Rightarrow T$, satisfyingAn algebra for a monad is $(A,a:TA\to A)$ with $a{\eta }_A=1_A$ and $aTa=a{\mu }_A$. A morphism of algebras for a monad $h:(A,a)\to (B,b)$ satisfies $ha=bTh$. These objects and arrows form the Eilenberg-Moore category ${\mathcal{C}}^T$.

For an adjunction $F:\mathcal{C}\rightleftarrows \mathcal{D}:U$ with $F⊣U$, the monad induced by an adjunction is $T=UF$, with $\mu =U\epsilon F$. The Eilenberg-Moore comparison functor is $K(D)=(UD,U{\epsilon }_D)$ and $K(h)=Uh$. The functor $U$ is monadic when this comparison is an equivalence of categories; this is a monadic adjunction.

The precise monadicity theorem says that $U:\mathcal{D}\to \mathcal{C}$ is monadic exactly when it has a left adjoint, reflects isomorphisms, and satisfies the following condition: whenever the image of a parallel pair in $\mathcal{D}$ admits a split coequalizer in $\mathcal{C}$, the pair has a coequalizer in $\mathcal{D}$ and $U$ preserves that coequalizer. Equivalently, these coequalizers are created up to the canonical isomorphism identifying their underlying coequalizer objects. Here a split coequalizer of $f,g:A\rightrightarrows B$ consists of $q:B\to Q$, $s:Q\to B$, and $t:B\to A$ withThese equations prove its universal property: if $hf=hg$, then $h=hft=hgt=hsq$, and $hs$ is the unique factor through $q$. Every functor preserves these equations, so every functor preserves split coequalizers.

First consider the forgetful functor $U^T:{\mathcal{C}}^T\to \mathcal{C}$. It has the free algebra functor $A\mapsto (TA,{\mu }_A)$ as a left adjoint: the inverse transposition formulas are $h\mapsto h{\eta }_A$ and $v\mapsto bTv$. An monad algebra morphism whose underlying arrow is invertible has an algebra-morphism inverse, by rearranging its structure equation. Thus $U^T$ reflects isomorphisms.

Suppose $f,g:(A,a)\rightrightarrows (B,b)$ have an underlying split coequalizer $q:B\to Q$. Applying $T$ gives a coequalizer of $Tf,Tg$. The equation $qbTf=qfa=qga=qbTg$ therefore gives a unique $c:TQ\to Q$ with $cTq=qb$. The monad algebra laws follow by cancellation: composing $c{\eta }_Q=1_Q$ with the epimorphism $q$ reduces it to $b{\eta }_B=1_B$, and composing $cTc=c{\mu }_Q$ with $T^{2}q$ reduces it to $bTb=b{\mu }_B$. Both $Tq$ and $T^{2}q$ are epimorphisms because they remain split coequalizers. Thus $(Q,c)$ is an algebra for a monad and $q$ is an monad algebra morphism. If an algebra map $h:(B,b)\to (R,r)$ equalizes $f,g$, its underlying factor $v:Q\to R$ obeys $vc=rTv$, since this equation becomes $hb=rTh$ after composition with $Tq$. This proves that $U^T$ creates the required coequalizers. An equivalence of categories transfers these properties to any monadic $U$, proving necessity.

Conversely assume the three conditions and write $T=UF$. Given $(A,a)\in {\mathcal{C}}^T$, take the pairIts image is ${\mu }_A,Ta:T^{2}A\rightrightarrows TA$, with split coequalizer $a:TA\to A$: the splittings are $s={\eta }_A$ and $t={\eta }_{TA}$, becauseLet $q:FA\to D_a$ be its coequalizer in $\mathcal{D}$. Preservation gives a canonical isomorphism $U{D}_a\cong A$ under which $Uq=a$. Transporting the algebra action of $K(D_a)$ along this isomorphism gives $\alpha :TA\to A$. Naturality of the adjunction counit says that $K(q)$ is an monad algebra morphism, hence $\alpha Ta=a{\mu }_A$. But $aTa=a{\mu }_A$ too, and $Ta$ is a split epimorphism, with section $T{\eta }_A$. Therefore $\alpha =a$. Every algebra is consequently isomorphic to some $K(D_a)$.

It remains to establish full faithfulness, rather than merely construct objects. For $B\in \mathcal{D}$, put $b=U{\epsilon }_B$. The adjunction counit ${\epsilon }_B:FUB\to B$ coequalizes ${\epsilon }_{FUB},Fb$. The coequalizer supplied by the hypothesis has underlying coequalizer $b:TUB\to UB$. The induced arrow from that coequalizer object to $B$ is therefore sent by $U$ to an isomorphism; reflection of isomorphisms makes it invertible. Thus ${\epsilon }_B$ itself is this coequalizer.

Now let $h:K(B)\to K(C)$ be an monad algebra morphism, and put $c=U{\epsilon }_C$. The arrow $v={\epsilon }_{C}Fh:FUB\to C$ coequalizes the two arrows in the preceding presentation. To check this directly, use transposition along $F⊣U$: the transposes of $vFb$ and $v{\epsilon }_{FUB}$ are respectively $hb$ and $cTh$, which agree by the algebra-morphism equation. Hence a unique $\bar{h}:B\to C$ satisfies $\bar{h}{\epsilon }_B=v$. Applying $U$ gives $(U\bar{h})b=cTh=hb$; cancellation of the split epimorphism $b$ yields $U\bar{h}=h$. Finally, any arrow $r:B\to C$ satisfies $r{\epsilon }_B={\epsilon }_{C}FUr$ by naturality, so two arrows with equal image under $U$ are equal by the coequalizer property. The comparison is full and faithful and essentially surjective, hence an equivalence of categories. This proves the theorem.

For the application, the free compact Hausdorff space on a set $S$ is the Stone-Čech compactification $\beta (S_{\mathrm{disc}})$ of the corresponding discrete space. Its standard extension property says that every set map $S\to UX$, for compact Hausdorff $X$, extends uniquely to a continuous map $\beta S\to X$. This constructs the required left adjoint. Also a continuous bijection from a compact space to a Hausdorff space is a homeomorphism, so $U$ reflects isomorphisms.

Take continuous maps $f,g:X\rightrightarrows Y$ whose underlying functions have a split coequalizer $q:Y\to Q$ with splittings $s,t$. The splittings need not be continuous. LetThe set $E$ is compact and hence closed. Since $qf=qg$, it is contained in $R$. Conversely $(y,sq(y))=(ft(y),gt(y))\in E$. ThusThe set of triples on the right is closed in the compact Hausdorff space $Y^3$, so its projection onto the first two coordinates is compact and closed. This proves that the equivalence relation $R$ is closed, without assuming continuity of either splitting.

Give $Q\cong Y/R$ the quotient topology. The allowed closed-relation criterion makes it a compact Hausdorff space. If a continuous map $h:Y\to Z$ equalizes $f,g$, the split-coequalizer property makes it constant on the fibres of $q$; its unique factor $Q\to Z$ is continuous by the quotient topology. Hence this is a coequalizer in compact Hausdorff spaces, preserved by $U$. Its topology is forced uniquely, since any continuous surjection from a compact space to a Hausdorff space is a quotient map. All three conditions hold, and therefore the forgetful functor from compact Hausdorff spaces to sets is monadic.

The smooth h-cobordism theorem states: if $(W;M_0,M_1)$ is a compact smooth h-cobordism, $\dim W\ge 6$, and its connected boundary manifolds are simply connected, then there is a diffeomorphism $W\cong M_0\times [0,1]$ restricting to the prescribed identification on $M_0$. In particular $M_0$ and $M_1$ are diffeomorphic. Being an h-cobordism means that both inclusions $M_i\hookrightarrow W$ are homotopy equivalences; their fundamental groups then also identify with that of $W$.

The homotopy-equivalence condition is necessary for a trivial cobordism: each end of a cylinder is a deformation retract, by $(x,t)\mapsto (x,(1-s)t)$ for the incoming end and $(x,t)\mapsto (x,(1-s)t+s)$ for the outgoing end. Its relative homology with respect to either end is consequently zero. Compactness is also necessary for a product with a closed compact end. The dimensional bound and simple connectivity are not necessary for an individual cobordism to be a product. The cylinder $S^2\times [0,1]$ is a simply connected trivial cobordism of dimension three; and, in the theorem's dimension range, $(S^1\times S^4)\times [0,1]$ is a six-dimensional trivial cobordism with fundamental group $\mathbb{Z}$. These prove the two non-necessity assertions explicitly, while the inclusion argument proves necessity of the h-condition.

For the remaining claims, use a cobordism Morse function: its two regular boundary levels are zero and one, and all critical points are interior. An arbitrary function with unrelated boundary values would not have the relative handle-count property. The relative Morse chain complex for $(W,M_0)$ has one generator per critical point, in its Morse index. ThusA trivial cobordism has zero relative homology, hence zero relative Euler characteristic. Modulo two the displayed sum is the number of critical points. An odd number makes it odd, and therefore nonzero. This proves an odd-critical-point cobordism is not trivial.

With precisely two critical points of indices $\lambda \le \mu$, a trivial cobordism requires the two-generator relative complex to be acyclic over $\mathbb{Z}$. Equal indices leave a rank-two homology group; a gap of more than one leaves both generators with zero possible differential. Therefore $\mu =\lambda +1$ is necessary. The remaining two-term complex isChoose a transverse gradient-like vector field. A nonzero differential requires the index-$\mu$ point to be above the index-$\lambda$ point; otherwise no descending connecting trajectories exist. In a common regular level between them, the belt sphere $S_U(p)$ has dimension $d-\lambda -1$, and the attaching sphere $S_L(q)$ has dimension $\mu -1=\lambda$, where $d=\dim W$. Their dimensions add to $d-1$, the level's dimension, so their transverse intersection consists of points. Each intersection is one descending trajectory from $q$ to $p$, modulo time translation. Orient the attaching sphere and use an orientation of a vector space on the handle core to coorient the belt sphere; the resulting signed count is the cellular boundary coefficientThe two homology groups are the kernel and cokernel of multiplication by $a$. They vanish precisely when $a$ is a unit of $\mathbb{Z}$. Consequentlyare necessary. This proves the two-critical-point obstruction to a trivial cobordism; it makes an algebraic incidence statement and does not assume that an algebraic count of one automatically supplies geometric handle cancellation.

For an explicit nontrivial cobordism from $S^3$ to itself, attach two zero-framed two-handles to the outgoing end of $S^3\times [0,1]$ along the standard Hopf link. Its two components are unknots with linking number one. This realizes two successive $(2,2)$ surgeries, as suggested by the hint. The outgoing boundary is again $S^3$, and this can be checked geometrically: the Hopf-link exterior is $T^2\times [0,1]$, and each component's zero-framing longitude is the other's meridian. Filling along these two slopes glues two solid tori with meridians intersecting once, the standard genus-one Heegaard decomposition of $S^3$. Thus zero surgery on the Hopf link does return the three-sphere, not just an unspecified homology sphere.

The trace has only the two added index-two handles. Its relative cellular complex has $C_2(W,S^3)={\mathbb{Z}}^2$ and all other groups zero, soIt cannot be a product cylinder. Equivalently this trace is $S^2\times S^2$ with two disjoint open four-balls removed: the usual handle presentation of $S^2\times S^2$ has one zero-handle, the two zero-framed Hopf-linked two-handles, and one four-handle. Removing the bottom and top balls exposes the two three-sphere ends. The relative-homology calculation alone already proves the requested cobordism is nontrivial.

Choose real constants $a_0<a_1<\cdots <a_m$ and define the smooth homogeneous quotientIt is unchanged by nonzero rescaling, so it is well defined on Real projective space. On the unit sphere its differential vanishes on the tangent space precisely when $Ax$ is proportional to $x$, where $A=\mathrm{diag}(a_0,\dots ,a_m)$. The entries are distinct, so the only critical projective lines are $p_i=[e_i]$. In the affine chart $x_i=1$, write $u_j=x_j/x_i$ for $j\ne i$. ThenIts Hessian at zero is diagonal with entries $2(a_j-a_i)$. Exactly $i$ are negative, and none vanish. Thus this is a Morse function with one critical point $p_i$ of index $i$ for every $0\le i\le m$.

Choose the quotient of the round metric on the unit sphere. The downward gradient flow has sphere equation ${\dot{x}}_j=-2(a_j-f(x))x_j$ and projective solutionNormalization to the unit sphere supplies the common positive factor. As $t\to +\infty$, the smallest index with nonzero coordinate dominates; as $t\to -\infty$, the largest dominates. ConsequentlyThese are its unstable manifolds and stable manifolds. If $j\ge k$, the tangent directions in $W^u(p_j)$ allow coordinates $0,\dots ,j$, and those in $W^s(p_k)$ allow $k,\dots ,m$, modulo the common projective line. Together they span the full tangent space at every intersection. If $j<k$ the intersection is empty. Thus the flow is a Morse-Smale gradient flow, with no generic perturbation needed.

Orient each unstable manifold using its affine coordinates. The Morse-Smale complex has $C_i=\mathbb{Z}{p}_i$; its differential counts oriented unparametrized connecting trajectories of index difference one. A trajectory from $p_i$ to $p_{i-1}$ has only coordinates $x_{i-1},x_i$ nonzero, and there are exactly two, distinguished by the sign of $x_{i-1}/x_i$. Each sign branch is a single orbit modulo time translation.

To calculate their signs rather than merely count two, use the characteristic closed disk of the unstable cell: the closed upper hemisphere of $S^i$ with $x_i\ge 0$, followed by projection onto ${\mathbb{RP}}^i$. Its interior is $W^u(p_i)$, and its boundary map is the antipodal double cover $S^{i-1}\to {\mathbb{RP}}^{i-1}$. After collapsing ${\mathbb{RP}}^{i-2}$, its degree on the $(i-1)$-cell is the sum of the degrees of the two sheets. Comparing the sheets by the antipodal map on $S^{i-1}$ gives relative orientation $(-1{)}^i$: it is induced by $-I$ on ${\mathbb{R}}^i$, whose determinant is $(-1{)}^i$. Thus the two trajectory signs add to $1+(-1{)}^i$, up to the choice of generator orientations. The projective quadratic Morse complex is thereforeIn particular ${\partial }_1=0$, as required for the connected one-dimensional skeleton. The signed trajectory count agrees with this attaching-map incidence calculation, so it computes integral rather than only mod-two homology.