Adjoint Functors and Freyd's Adjoint Functor Theorem

💡 The Core Insight

An adjunction $F \dashv G$ means $F$ provides the "best $\mathcal{D}$ -approximation" to objects in $\mathcal{C}$ , while $G$ provides the "best $\mathcal{C}$ -approximation" to objects in $\mathcal{D}$ . Freyd's theorem says that if a functor preserves all limits and satisfies a solution set condition, it automatically has a left adjoint—the universal construction exists without explicit construction.

Definition

Adjoint Functors: Let $F: \mathcal{C} \to \mathcal{D}$ and $G: \mathcal{D} \to \mathcal{C}$ be functors. We say $F$ is left adjoint to $G$ (written $F \dashv G$ ) if there exists a natural isomorphism

\Phi_{c,d}: \text{Hom}_{\mathcal{D}}(F(c), d) \xrightarrow{\sim} \text{Hom}_{\mathcal{C}}(c, G(d))

for all objects $c \in \mathcal{C}$ and $d \in \mathcal{D}$ . Equivalently, for each $c \in \mathcal{C}$ there exists a unit morphism $\eta_c: c \to G(F(c))$ universal in the sense that for any $f: c \to G(d)$ there exists a unique $\bar{f}: F(c) \to d$ with $f = G(\bar{f}) \circ \eta_c$ .

Freyd's Adjoint Functor Theorem (GAFT): Let $G: \mathcal{D} \to \mathcal{C}$ be a functor where $\mathcal{D}$ is locally small and complete. Then $G$ has a left adjoint if and only if: 1. $G$ preserves all (small) limits 2. $G$ satisfies the solution set condition: for each $c \in \mathcal{C}$ , there exists a set $S$ of morphisms $\{f_i: c \to G(d_i)\}_{i \in I}$ such that every morphism $g: c \to G(d)$ factors as $g = G(h) \circ f_i$ for some $i \in I$ and $h: d_i \to d$ .

TheoremMain Result

Theorem 1 (Freyd's General Adjoint Functor Theorem): Let $\mathcal{C}$ and $\mathcal{D}$ be locally small categories with $\mathcal{D}$ complete. A functor $G: \mathcal{D} \to \mathcal{C}$ has a left adjoint if and only if $G$ preserves limits and satisfies the solution set condition.

Theorem 2 (Special Adjoint Functor Theorem): Let $\mathcal{C}$ be a locally presentable category and $\mathcal{D}$ a cocomplete category. A functor $G: \mathcal{D} \to \mathcal{C}$ that preserves all colimits has a left adjoint if and only if it satisfies the solution set condition. If $\mathcal{C}$ is complete, well-powered, and has a cogenerator, then $G$ preserving limits automatically guarantees the solution set condition.

Proof

We prove Theorem 1. The "only if" direction is straightforward: if $F \dashv G$ , then $G$ preserves limits because $\text{Hom}_{\mathcal{C}}(c, G(\lim d_j)) \cong \text{Hom}_{\mathcal{D}}(F(c), \lim d_j) \cong \lim \text{Hom}_{\mathcal{D}}(F(c), d_j) \cong \lim \text{Hom}_{\mathcal{C}}(c, G(d_j)) \cong \text{Hom}_{\mathcal{C}}(c, \lim G(d_j))$ by Yoneda lemma. The solution set exists by taking a singleton containing any component of the unit. For the "if" direction, fix $c \in \mathcal{C}$ . We construct $F(c)$ as a limit. Let $S = \{f_i: c \to G(d_i)\}_{i \in I}$ be a solution set for $c$ . Consider the comma category $(c \downarrow G)$ whose objects are pairs $(d, f: c \to G(d))$ and morphisms from $(d, f)$ to $(d', f')$ are arrows $h: d \to d'$ in $\mathcal{D}$ with $G(h) \circ f = f'$ . Define the projection functor $\pi: (c \downarrow G) \to \mathcal{D}$ sending $(d, f) \mapsto d$ . Since $\mathcal{D}$ is complete, the limit $L = \lim \pi$ exists in $\mathcal{D}$ . We claim $F(c) = L$ with universal morphism given by the composition

\eta_c: c \xrightarrow{\Delta} \lim_{(d,f) \in (c \downarrow G)} G(d) \xrightarrow{G(\lim \text{proj})} G(L)

where the first arrow is the diagonal map and the second uses that $G$ preserves limits. To verify universality, let $g: c \to G(d')$ be any morphism. By the solution set condition, there exists $i \in I$ and $h: d_i \to d'$ with $g = G(h) \circ f_i$ . This shows $(d', g)$ is in the comma category $(c \downarrow G)$ , giving a projection $p_{d',g}: L \to d'$ from the limit. We claim $\bar{g} = p_{d',g}$ satisfies $g = G(\bar{g}) \circ \eta_c$ . Indeed, since $G$ preserves limits, $G(L) \cong \lim_{(d,f) \in (c \downarrow G)} G(d)$ , and the component of this limit at $(d', g)$ is precisely $G(p_{d',g})$ . The definition of $\eta_c$ and the universal property of the limit ensure that the component at $(d', g)$ of $G(\text{lim proj}) \circ \Delta$ is exactly $g$ , so $G(p_{d',g}) \circ \eta_c = g$ . For uniqueness, suppose $\bar{g}': F(c) \to d'$ also satisfies $g = G(\bar{g}') \circ \eta_c$ . The morphism $\bar{g}'$ induces a cone from $L$ to the diagram $\pi$ . By the universal property of the limit, this cone is determined by its components, which are in turn determined by composing with $G$ . Since $G(\bar{g}') \circ \eta_c = g = G(\bar{g}) \circ \eta_c$ and $\eta_c$ is the universal cone to $G(L)$ , we have $\bar{g}' = \bar{g}$ . To extend this to a functor $F: \mathcal{C} \to \mathcal{D}$ and verify the adjunction is natural, we use similar limit constructions for morphisms in $\mathcal{C}$ and apply functoriality of limits. □

∎

⚠️ Common mistake

Common Error: The solution set condition is not automatically satisfied even when $G$ preserves limits! The category of sets is complete, and the identity functor preserves all limits, yet in certain large categories without cardinality bounds, the solution set condition can fail. Also, students often confuse which functor needs which property: it's the right adjoint $G$ that must preserve limits, not the left adjoint (which preserves colimits). Finally, the solution set must be an actual set, not a proper class—this is a genuine size restriction that can fail in pathological categories.

mathematician

Peter Freyd (1964) — Proved the General Adjoint Functor Theorem in his groundbreaking thesis "Abelian Categories," providing the first general criterion for existence of adjoint functors and founding much of modern categorical algebra.

visualization

Example1

Free Group Functor: Let $U: \mathbf{Grp} \to \mathbf{Set}$ be the forgetful functor that sends each group to its underlying set. We show $U$ has a left adjoint $F: \mathbf{Set} \to \mathbf{Grp}$ , the free group functor, using Freyd's theorem. First, $\mathbf{Grp}$ is complete (products of groups are constructed componentwise, equalizers are kernels). Second, $U$ preserves limits: the underlying set of a product of groups is the product of underlying sets, and similarly for equalizers. For the solution set condition, fix a set $X$ . Consider the set $S = \{\iota: X \to U(F(X))\}$ where $\iota$ is the canonical inclusion and $F(X)$ is the free group on $X$ (which we construct explicitly). For any homomorphism $f: X \to U(G)$ where $G$ is a group, the universal property of free groups gives a unique homomorphism $\bar{f}: F(X) \to G$ with $f = U(\bar{f}) \circ \iota$ . Thus $S$ is a solution set (a singleton!). By Freyd's theorem, $U$ has a left adjoint. The construction in the proof actually recovers the free group: for a set $X$ , the free group $F(X)$ consists of reduced words in $X \cup X^{-1}$ , and the unit $\eta_X: X \to U(F(X))$ sends each element to the corresponding word of length 1.

Example2

Sheafification: In sheaf theory, let $\mathcal{C}$ be the category of presheaves on a topological space $X$ with values in $\mathbf{Set}$ , and let $\mathcal{D}$ be the full subcategory of sheaves. The inclusion functor $i: \mathcal{D} \to \mathcal{C}$ has a left adjoint called sheafification. We verify this using Freyd's theorem. The category $\mathcal{D}$ is complete (limits of sheaves are computed as in presheaves, and the sheaf condition is preserved). The inclusion $i$ preserves limits trivially since it's a full inclusion and limits are computed the same way. For the solution set condition, fix a presheaf $\mathcal{F}$ . Consider the étalé space construction $\text{Et}(\mathcal{F})$ , which gives a sheaf $\mathcal{F}^+$ , together with the natural morphism $\theta: \mathcal{F} \to i(\mathcal{F}^+)$ defined by taking germs. The set $S = \{\theta\}$ is a solution set: if $\phi: \mathcal{F} \to i(\mathcal{G})$ for any sheaf $\mathcal{G}$ , then since $\mathcal{G}$ already satisfies the sheaf condition, $\phi$ factors uniquely through $\theta$ by the universal property of taking stalks and sections. The sheafification functor $(-)^+: \mathcal{C} \to \mathcal{D}$ is the left adjoint guaranteed by Freyd's theorem, giving the natural bijection $\text{Hom}_{\mathcal{D}}(\mathcal{F}^+, \mathcal{G}) \cong \text{Hom}_{\mathcal{C}}(\mathcal{F}, i(\mathcal{G}))$ .

Counterexample

Failure without Solution Set Condition: Consider the category $\mathbf{Set}$ and let $\kappa$ be an inaccessible cardinal. Define $\mathcal{C}$ to be the full subcategory of $\mathbf{Set}$ consisting of sets of cardinality $< \kappa$ , and $\mathcal{D} = \mathbf{Set}$ . Let $G: \mathcal{D} \to \mathcal{C}$ be the functor that sends each set $X$ to $X$ if $|X| < \kappa$ and to $\emptyset$ otherwise. The functor $G$ preserves all limits (limits decrease or preserve cardinality in $\mathbf{Set}$ , so limits in $\mathcal{D}$ of sets smaller than $\kappa$ remain smaller than $\kappa$ ). However, $G$ does not have a left adjoint. The issue is the solution set condition. Consider the singleton set $\{*\} \in \mathcal{C}$ . For $G$ to have a left adjoint, we'd need a "free" object $F(\{*\})$ and a universal map $\eta: \{*\} \to G(F(\{*\}))$ . But for any set $X$ with $|X| \geq \kappa$ , we have $G(X) = \emptyset$ , so there can be no morphism from $\{*\}$ to $G(X)$ factoring through the unit. For sets $X$ with $|X| < \kappa$ , the morphisms $\{*\} \to X$ are just elements of $X$ , and there are unboundedly many such sets below $\kappa$ (since $\kappa$ is inaccessible). No set of morphisms can serve as a solution set because we'd need representatives for arbitrary large cardinalities below $\kappa$ , which would form a proper class, not a set.

🧠 Intuition

Think of an adjunction $F \dashv G$ as an "optimal approximation" relationship. If $G: \mathcal{D} \to \mathcal{C}$ "forgets" structure (like the forgetful functor from groups to sets), then its left adjoint $F$ "freely adds" that structure back in the most economical way. The unit $\eta_c: c \to GF(c)$ embeds your object into the underlying object of its free version. Geometrically, imagine $\mathcal{C}$ as a "simple" space and $\mathcal{D}$ as a "structured" space. The right adjoint $G$ projects from structured to simple. The left adjoint $F$ lifts from simple to structured, choosing the "freest" or "most general" structured object that maps back down to your simple object. Freyd's theorem says: if $G$ respects all limits (it's "continuous" in a categorical sense) and there's only a set's worth of "essentially different" ways to approximate any object (solution set condition), then the optimal approximations $F(c)$ exist automatically as certain limits. You don't need to construct them by hand—the theorem guarantees their existence through abstract nonsense.

connection

Stone Duality and Categorical Logic: The equivalence between Boolean algebras and Stone spaces (compact Hausdorff zero-dimensional spaces) extends to a profound connection between adjoint functors and logical theories. The Lindenbaum-Tarski algebra construction is left adjoint to the functor taking a Boolean algebra to its theory (set of provable formulas). Freyd's theorem applies to show that syntax-semantics adjunctions exist systematically: the functor from models to theories (taking a model to its complete theory) has a left adjoint (taking a theory to its free model or term algebra) precisely when the theory is sufficiently "small." This bridges categorical logic, model theory, and theoretical computer science, where such adjunctions underlie the Curry-Howard correspondence and the semantics of programming languages.

Adjoint Functors and Freyd's Adjoint Functor Theorem

Adjoint Functors and Freyd's Adjoint Functor Theorem

Overview

Core Theory

Examples and Applications

Historical Development

Connections to Other Areas

What You Need First

What This Opens Up

See Also