Equivalence of Categories Induces an Adjoint Equivalence (Theorem # 4155)
Theorem
Let $\mathcal C$ and $\mathcal D$ be categories, and let $F: \mathcal C \to \mathcal D$ and $G: \mathcal D \to \mathcal C$ be functors.
If $F$ and $G$ form an equivalence of categories, meaning that there exist natural isomorphisms $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ and $\delta: FG \Rightarrow \operatorname{id}_{\mathcal D}$, then $\eta$ and $\delta$ may be replaced, if necessary, by natural isomorphisms
\begin{align*}
\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF,
\qquad
\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}
\end{align*}
such that $F \dashv G$ with unit $\eta$ and counit $\varepsilon$. In particular, both the unit and counit of this adjunction are natural isomorphisms.
Conversely, if $F \dashv G$ is an adjunction whose unit $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ and counit $\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}$ are natural isomorphisms, then $F$ and $G$ form an equivalence of categories.
Discussion
This result records Equivalence of Categories Induces an Adjoint Equivalence: given C and D be categories, and let F: C D and G: D C be functors. If F and G form an equivalence of categories, meaning that there exist natural isomorphisms : id C GF and : FG id D, then and may be replaced, if.... It is used to organize adjunctions and universal constructions and to make later constructions depend only on the stated universal or exactness property.
Proof
[proofplan]
The converse direction is formal: an adjunction whose unit and counit are natural isomorphisms already supplies the two natural isomorphisms required in the definition of equivalence. For the forward direction, we start from arbitrary equivalence data $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ and $\delta: FG \Rightarrow \operatorname{id}_{\mathcal D}$, keep $\eta$ fixed, and modify the counit to a new natural isomorphism $\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}$. The definition of $\varepsilon$ is chosen so that the triangle identity for $G$ holds by construction; naturality of $\eta$ and $\delta$ then gives the triangle identity for $F$. Once the two triangle identities are established, the usual hom-set formulas define the adjunction $F \dashv G$.
[/proofplan]
[step:Extract the equivalence data and define the adjusted counit]
Assume first that $F$ and $G$ form an equivalence of categories. Thus there are natural isomorphisms
\begin{align*}
\eta: \operatorname{id}_{\mathcal C} &\Rightarrow GF,\\
\delta: FG &\Rightarrow \operatorname{id}_{\mathcal D}.
\end{align*}
For each object $d \in \mathcal D$, define a morphism
\begin{align*}
\varepsilon_d: FGd \to d
\end{align*}
by
\begin{align*}
\varepsilon_d
:=
\delta_d \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}.
\end{align*}
The types are as follows:
\begin{align*}
FGd
\xrightarrow{\delta_{FGd}^{-1}}
FGFGd
\xrightarrow{F(\eta_{Gd}^{-1})}
FGd
\xrightarrow{\delta_d}
d.
\end{align*}
Since $\delta_{FGd}$, $\eta_{Gd}$, and $\delta_d$ are isomorphisms, each $\varepsilon_d$ is an isomorphism.
[guided]
We begin with an equivalence of categories. This means that $F$ and $G$ are already quasi-inverse functors, but the given natural isomorphisms need not satisfy the triangle identities required of an adjunction. We keep the natural isomorphism
\begin{align*}
\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF
\end{align*}
as the prospective unit.
The given comparison
\begin{align*}
\delta: FG \Rightarrow \operatorname{id}_{\mathcal D}
\end{align*}
is not necessarily the right counit. We therefore define a new family of morphisms. For every object $d \in \mathcal D$, set
\begin{align*}
\varepsilon_d
:=
\delta_d \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}.
\end{align*}
This formula is type-correct because
\begin{align*}
\delta_{FGd}^{-1}&: FGd \to FGFGd,\\
F(\eta_{Gd}^{-1})&: FGFGd \to FGd,\\
\delta_d&: FGd \to d.
\end{align*}
Thus the composite is a morphism $\varepsilon_d: FGd \to d$. Each factor is an isomorphism, so $\varepsilon_d$ is also an isomorphism.
[/guided]
[/step]
[step:Verify that the adjusted counit is natural]
Let $h: d \to d'$ be a morphism in $\mathcal D$. We prove
\begin{align*}
h \circ \varepsilon_d
=
\varepsilon_{d'} \circ FGh.
\end{align*}
Using the definition of $\varepsilon_d$ and naturality of $\delta$ at $h$, naturality of $\delta$ at $FGh$, and naturality of $\eta^{-1}$ at $Gh$, we compute
\begin{align*}
h \circ \varepsilon_d
&=
h \circ \delta_d \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ FGh \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(Gh \circ \eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1} \circ GFGh) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1}) \circ FGFGh \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1}) \circ \delta_{FGd'}^{-1} \circ FGh\\
&=
\varepsilon_{d'} \circ FGh.
\end{align*}
Hence $\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}$ is a natural isomorphism.
[guided]
To prove that $\varepsilon$ is natural, we must show that every morphism $h: d \to d'$ in $\mathcal D$ makes the square for $\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}$ commute. Equivalently, we must prove
\begin{align*}
h \circ \varepsilon_d
=
\varepsilon_{d'} \circ FGh.
\end{align*}
Substitute the definition of $\varepsilon_d$:
\begin{align*}
h \circ \varepsilon_d
&=
h \circ \delta_d \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}.
\end{align*}
Naturality of $\delta$ applied to $h: d \to d'$ gives
\begin{align*}
h \circ \delta_d
=
\delta_{d'} \circ FGh.
\end{align*}
Therefore
\begin{align*}
h \circ \varepsilon_d
&=
\delta_{d'} \circ FGh \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(Gh \circ \eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}.
\end{align*}
Naturality of the inverse natural transformation $\eta^{-1}: GF \Rightarrow \operatorname{id}_{\mathcal C}$ applied to $Gh: Gd \to Gd'$ gives
\begin{align*}
Gh \circ \eta_{Gd}^{-1}
=
\eta_{Gd'}^{-1} \circ GFGh.
\end{align*}
Using this identity inside $F$ gives
\begin{align*}
h \circ \varepsilon_d
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1} \circ GFGh) \circ \delta_{FGd}^{-1}\\
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1}) \circ FGFGh \circ \delta_{FGd}^{-1}.
\end{align*}
Finally, naturality of $\delta$ applied to $FGh: FGd \to FGd'$ gives
\begin{align*}
FGh \circ \delta_{FGd}
=
\delta_{FGd'} \circ FGFGh.
\end{align*}
Since each component of $\delta$ is an isomorphism, this is equivalent to
\begin{align*}
FGFGh \circ \delta_{FGd}^{-1}
=
\delta_{FGd'}^{-1} \circ FGh.
\end{align*}
Substituting this into the previous composite yields
\begin{align*}
h \circ \varepsilon_d
&=
\delta_{d'} \circ F(\eta_{Gd'}^{-1}) \circ \delta_{FGd'}^{-1} \circ FGh\\
&=
\varepsilon_{d'} \circ FGh.
\end{align*}
Thus $\varepsilon$ is natural. Since each component $\varepsilon_d$ is already known to be an isomorphism, $\varepsilon$ is a natural isomorphism.
[/guided]
[/step]
[step:Prove the triangle identity for $G$]
For each object $d \in \mathcal D$, we prove
\begin{align*}
G\varepsilon_d \circ \eta_{Gd}
=
\operatorname{id}_{Gd}.
\end{align*}
Naturality of $\delta$ applied to the morphism $\delta_d^{-1}: d \to FGd$ gives
\begin{align*}
\delta_d^{-1} \circ \delta_d
=
\delta_{FGd} \circ FG(\delta_d^{-1}).
\end{align*}
Since $\delta_d^{-1} \circ \delta_d = \operatorname{id}_{FGd}$, this gives
\begin{align*}
FG(\delta_d^{-1})
=
\delta_{FGd}^{-1}.
\end{align*}
Applying $G$ yields
\begin{align*}
GFG(\delta_d^{-1})
=
G(\delta_{FGd}^{-1}).
\end{align*}
Naturality of $\eta$ applied to the morphism $G(\delta_d^{-1}): Gd \to GFGd$ gives
\begin{align*}
GFG(\delta_d^{-1}) \circ \eta_{Gd}
=
\eta_{GFGd} \circ G(\delta_d^{-1}).
\end{align*}
Naturality of $\eta$ applied to the morphism $\eta_{Gd}^{-1}: GFGd \to Gd$ gives
\begin{align*}
GF(\eta_{Gd}^{-1}) \circ \eta_{GFGd}
=
\eta_{Gd} \circ \eta_{Gd}^{-1}
=
\operatorname{id}_{GFGd}.
\end{align*}
Applying $G$ to the definition of $\varepsilon_d$ and composing with $\eta_{Gd}$ gives
\begin{align*}
G\varepsilon_d \circ \eta_{Gd}
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ G(\delta_{FGd}^{-1}) \circ \eta_{Gd}\\
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ GFG(\delta_d^{-1}) \circ \eta_{Gd}\\
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ \eta_{GFGd} \circ G(\delta_d^{-1})\\
&=
G(\delta_d) \circ G(\delta_d^{-1})\\
&=
\operatorname{id}_{Gd}.
\end{align*}
Thus the $G$-triangle identity holds.
[guided]
The delicate point is that $\eta$ is a natural transformation on $\mathcal C$, so it may only be applied to morphisms in $\mathcal C$. The morphism $\delta_d: FGd \to d$ lies in $\mathcal D$, so we first convert it into a morphism in $\mathcal C$ by applying $G$ to the inverse $\delta_d^{-1}: d \to FGd$.
Naturality of $\delta: FG \Rightarrow \operatorname{id}_{\mathcal D}$ applied to $\delta_d^{-1}: d \to FGd$ says
\begin{align*}
\delta_d^{-1} \circ \delta_d
=
\delta_{FGd} \circ FG(\delta_d^{-1}).
\end{align*}
The left-hand side is $\operatorname{id}_{FGd}$ because $\delta_d$ is an isomorphism. Hence
\begin{align*}
FG(\delta_d^{-1})
=
\delta_{FGd}^{-1}.
\end{align*}
After applying the functor $G$, this becomes
\begin{align*}
GFG(\delta_d^{-1})
=
G(\delta_{FGd}^{-1}).
\end{align*}
Now apply naturality of $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ to the morphism $G(\delta_d^{-1}): Gd \to GFGd$ in $\mathcal C$. This gives
\begin{align*}
GFG(\delta_d^{-1}) \circ \eta_{Gd}
=
\eta_{GFGd} \circ G(\delta_d^{-1}).
\end{align*}
We also apply naturality of $\eta$ to the morphism $\eta_{Gd}^{-1}: GFGd \to Gd$. The result is
\begin{align*}
GF(\eta_{Gd}^{-1}) \circ \eta_{GFGd}
=
\eta_{Gd} \circ \eta_{Gd}^{-1}
=
\operatorname{id}_{GFGd}.
\end{align*}
These two naturality identities are exactly what the definition of $\varepsilon_d$ was designed to use.
Applying $G$ to
\begin{align*}
\varepsilon_d
=
\delta_d \circ F(\eta_{Gd}^{-1}) \circ \delta_{FGd}^{-1}
\end{align*}
and then composing with $\eta_{Gd}$ gives
\begin{align*}
G\varepsilon_d \circ \eta_{Gd}
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ G(\delta_{FGd}^{-1}) \circ \eta_{Gd}\\
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ GFG(\delta_d^{-1}) \circ \eta_{Gd}\\
&=
G(\delta_d) \circ GF(\eta_{Gd}^{-1}) \circ \eta_{GFGd} \circ G(\delta_d^{-1})\\
&=
G(\delta_d) \circ G(\delta_d^{-1})\\
&=
\operatorname{id}_{Gd}.
\end{align*}
Thus $G\varepsilon_d \circ \eta_{Gd}=\operatorname{id}_{Gd}$ for every object $d \in \mathcal D$, which is the triangle identity for $G$.
[/guided]
[/step]
[step:Prove the triangle identity for $F$]
For each object $c \in \mathcal C$, we prove
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
=
\operatorname{id}_{Fc}.
\end{align*}
By the definition of $\varepsilon$,
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1}) \circ \delta_{FGFc}^{-1} \circ F\eta_c.
\end{align*}
Naturality of $\delta$ applied to the morphism $F\eta_c: Fc \to FGFc$ gives
\begin{align*}
F\eta_c \circ \delta_{Fc}
=
\delta_{FGFc} \circ FGF\eta_c.
\end{align*}
After rearranging this identity through the displayed composite, it follows that
\begin{align*}
\delta_{FGFc}^{-1} \circ F\eta_c
=
FGF\eta_c \circ \delta_{Fc}^{-1}.
\end{align*}
Therefore
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1}) \circ FGF\eta_c \circ \delta_{Fc}^{-1}\\
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1} \circ GF\eta_c) \circ \delta_{Fc}^{-1}.
\end{align*}
Naturality of $\eta$ applied to the morphism $\eta_c: c \to GFc$ gives
\begin{align*}
GF\eta_c \circ \eta_c
=
\eta_{GFc} \circ \eta_c.
\end{align*}
Since $\eta_c$ is an isomorphism, cancellation on the right gives
\begin{align*}
GF\eta_c
=
\eta_{GFc}.
\end{align*}
Hence
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\operatorname{id}_{GFc}) \circ \delta_{Fc}^{-1}\\
&=
\operatorname{id}_{Fc}.
\end{align*}
Thus the $F$-triangle identity holds.
[guided]
We now prove the other triangle identity. For each object $c \in \mathcal C$, the desired equality is
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
=
\operatorname{id}_{Fc}.
\end{align*}
Substitute the definition of the adjusted counit at the object $Fc \in \mathcal D$:
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1}) \circ \delta_{FGFc}^{-1} \circ F\eta_c.
\end{align*}
The term to simplify is $\delta_{FGFc}^{-1} \circ F\eta_c$. Naturality of $\delta: FG \Rightarrow \operatorname{id}_{\mathcal D}$ applied to the morphism $F\eta_c: Fc \to FGFc$ gives
\begin{align*}
F\eta_c \circ \delta_{Fc}
=
\delta_{FGFc} \circ FGF\eta_c.
\end{align*}
Composing this equality on the left by $\delta_{FGFc}^{-1}$ and on the right by $\delta_{Fc}^{-1}$ gives
\begin{align*}
\delta_{FGFc}^{-1} \circ F\eta_c
=
FGF\eta_c \circ \delta_{Fc}^{-1}.
\end{align*}
Substituting into the displayed composite yields
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1}) \circ FGF\eta_c \circ \delta_{Fc}^{-1}\\
&=
\delta_{Fc} \circ F(\eta_{GFc}^{-1} \circ GF\eta_c) \circ \delta_{Fc}^{-1}.
\end{align*}
It remains to identify the middle composite in $\mathcal C$. Naturality of $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ applied to $\eta_c: c \to GFc$ gives
\begin{align*}
GF\eta_c \circ \eta_c
=
\eta_{GFc} \circ \eta_c.
\end{align*}
Because $\eta_c$ is an isomorphism, we may cancel it on the right and obtain
\begin{align*}
GF\eta_c
=
\eta_{GFc}.
\end{align*}
Therefore
\begin{align*}
\eta_{GFc}^{-1} \circ GF\eta_c
=
\operatorname{id}_{GFc}.
\end{align*}
Substitution gives
\begin{align*}
\varepsilon_{Fc} \circ F\eta_c
&=
\delta_{Fc} \circ F(\operatorname{id}_{GFc}) \circ \delta_{Fc}^{-1}\\
&=
\delta_{Fc} \circ \operatorname{id}_{FGFc} \circ \delta_{Fc}^{-1}\\
&=
\operatorname{id}_{Fc}.
\end{align*}
Thus $\varepsilon_{Fc} \circ F\eta_c=\operatorname{id}_{Fc}$ for every object $c \in \mathcal C$, which is the triangle identity for $F$.
[/guided]
[/step]
[step:Construct the adjunction from the two triangle identities]
For objects $c \in \mathcal C$ and $d \in \mathcal D$, define maps between hom-sets
\begin{align*}
\Phi_{c,d}: \operatorname{Hom}_{\mathcal D}(Fc,d) &\to \operatorname{Hom}_{\mathcal C}(c,Gd),\\
f &\mapsto Gf \circ \eta_c,
\end{align*}
and
\begin{align*}
\Psi_{c,d}: \operatorname{Hom}_{\mathcal C}(c,Gd) &\to \operatorname{Hom}_{\mathcal D}(Fc,d),\\
g &\mapsto \varepsilon_d \circ Fg.
\end{align*}
For $f: Fc \to d$,
\begin{align*}
\Psi_{c,d}(\Phi_{c,d}(f))
&=
\varepsilon_d \circ F(Gf \circ \eta_c)\\
&=
\varepsilon_d \circ FGf \circ F\eta_c\\
&=
f \circ \varepsilon_{Fc} \circ F\eta_c\\
&=
f.
\end{align*}
The third equality is naturality of $\varepsilon$ applied to $f: Fc \to d$, and the fourth equality is the $F$-triangle identity.
For $g: c \to Gd$,
\begin{align*}
\Phi_{c,d}(\Psi_{c,d}(g))
&=
G(\varepsilon_d \circ Fg) \circ \eta_c\\
&=
G\varepsilon_d \circ GFg \circ \eta_c\\
&=
G\varepsilon_d \circ \eta_{Gd} \circ g\\
&=
g.
\end{align*}
The third equality is naturality of $\eta$ applied to $g: c \to Gd$, and the fourth equality is the $G$-triangle identity. Hence $\Phi_{c,d}$ is a bijection with inverse $\Psi_{c,d}$.
We verify naturality in both variables. If $a: c' \to c$ is a morphism in $\mathcal C$ and $f: Fc \to d$ is a morphism in $\mathcal D$, then naturality of $\eta$ at $a$ gives
\begin{align*}
\Phi_{c',d}(f \circ Fa)
&=
G(f \circ Fa) \circ \eta_{c'}\\
&=
Gf \circ GFa \circ \eta_{c'}\\
&=
Gf \circ \eta_c \circ a\\
&=
\Phi_{c,d}(f) \circ a.
\end{align*}
If $b: d \to d'$ is a morphism in $\mathcal D$, then functoriality of $G$ gives
\begin{align*}
\Phi_{c,d'}(b \circ f)
&=
G(b \circ f) \circ \eta_c\\
&=
Gb \circ Gf \circ \eta_c\\
&=
Gb \circ \Phi_{c,d}(f).
\end{align*}
Thus $\Phi$ is natural in $c$ and $d$. These natural hom-set bijections are the standard hom-set characterization of an [adjunction](/page/Adjunction), with unit $\eta$ and counit $\varepsilon$ because the defining formulas are $f \mapsto Gf \circ \eta_c$ and $g \mapsto \varepsilon_d \circ Fg$. Therefore $F \dashv G$ with unit $\eta$ and counit $\varepsilon$, and both are natural isomorphisms.
[guided]
We now build the adjunction from the unit-counit data. For objects $c \in \mathcal C$ and $d \in \mathcal D$, define maps between hom-sets by
\begin{align*}
\Phi_{c,d}: \operatorname{Hom}_{\mathcal D}(Fc,d) &\to \operatorname{Hom}_{\mathcal C}(c,Gd),\\
f &\mapsto Gf \circ \eta_c,
\end{align*}
and
\begin{align*}
\Psi_{c,d}: \operatorname{Hom}_{\mathcal C}(c,Gd) &\to \operatorname{Hom}_{\mathcal D}(Fc,d),\\
g &\mapsto \varepsilon_d \circ Fg.
\end{align*}
We first prove that these maps are inverse bijections. Let $f: Fc \to d$ be a morphism in $\mathcal D$. Then
\begin{align*}
\Psi_{c,d}(\Phi_{c,d}(f))
&=
\varepsilon_d \circ F(Gf \circ \eta_c)\\
&=
\varepsilon_d \circ FGf \circ F\eta_c\\
&=
f \circ \varepsilon_{Fc} \circ F\eta_c\\
&=
f.
\end{align*}
The third equality is naturality of $\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D}$ applied to $f: Fc \to d$, and the fourth equality is the $F$-triangle identity.
Let $g: c \to Gd$ be a morphism in $\mathcal C$. Then
\begin{align*}
\Phi_{c,d}(\Psi_{c,d}(g))
&=
G(\varepsilon_d \circ Fg) \circ \eta_c\\
&=
G\varepsilon_d \circ GFg \circ \eta_c\\
&=
G\varepsilon_d \circ \eta_{Gd} \circ g\\
&=
g.
\end{align*}
The third equality is naturality of $\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF$ applied to $g: c \to Gd$, and the fourth equality is the $G$-triangle identity. Hence $\Phi_{c,d}$ and $\Psi_{c,d}$ are inverse bijections.
We must also verify that the bijection is natural in both variables. Let $a: c' \to c$ be a morphism in $\mathcal C$ and let $f: Fc \to d$ be a morphism in $\mathcal D$. Naturality in the $\mathcal C$-variable means compatibility with precomposition by $a$ on the right and by $Fa$ on the left. Using naturality of $\eta$ at $a$, we compute
\begin{align*}
\Phi_{c',d}(f \circ Fa)
&=
G(f \circ Fa) \circ \eta_{c'}\\
&=
Gf \circ GFa \circ \eta_{c'}\\
&=
Gf \circ \eta_c \circ a\\
&=
\Phi_{c,d}(f) \circ a.
\end{align*}
Now let $b: d \to d'$ be a morphism in $\mathcal D$. Naturality in the $\mathcal D$-variable means compatibility with postcomposition by $b$ and by $Gb$. By functoriality of $G$,
\begin{align*}
\Phi_{c,d'}(b \circ f)
&=
G(b \circ f) \circ \eta_c\\
&=
Gb \circ Gf \circ \eta_c\\
&=
Gb \circ \Phi_{c,d}(f).
\end{align*}
Thus the family $\Phi_{c,d}$ is natural in $c$ and $d$.
The standard hom-set characterization of an [adjunction](/page/Adjunction) says that natural bijections
\begin{align*}
\operatorname{Hom}_{\mathcal D}(Fc,d) \cong \operatorname{Hom}_{\mathcal C}(c,Gd)
\end{align*}
are equivalent to an adjunction $F \dashv G$. The formulas above show that the associated unit is $\eta$ and the associated counit is $\varepsilon$. Since $\eta$ was given as a natural isomorphism and $\varepsilon$ was constructed as a natural isomorphism, this adjunction is an adjoint equivalence.
[/guided]
[/step]
[step:Recover an equivalence from an adjunction with invertible unit and counit]
Conversely, suppose $F \dashv G$ has unit
\begin{align*}
\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF
\end{align*}
and counit
\begin{align*}
\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D},
\end{align*}
and suppose both $\eta$ and $\varepsilon$ are natural isomorphisms. These are exactly the natural isomorphisms required to exhibit $G$ as a quasi-inverse to $F$. Hence $F$ and $G$ form an equivalence of categories.
[guided]
For the converse direction, assume that $F \dashv G$ has unit
\begin{align*}
\eta: \operatorname{id}_{\mathcal C} \Rightarrow GF
\end{align*}
and counit
\begin{align*}
\varepsilon: FG \Rightarrow \operatorname{id}_{\mathcal D},
\end{align*}
and assume that both natural transformations are natural isomorphisms. To exhibit an equivalence of categories, we need functors $F: \mathcal C \to \mathcal D$ and $G: \mathcal D \to \mathcal C$ together with natural isomorphisms from $\operatorname{id}_{\mathcal C}$ to $GF$ and from $FG$ to $\operatorname{id}_{\mathcal D}$. The functors are the given functors, the first natural isomorphism is $\eta$, and the second natural isomorphism is $\varepsilon$. Therefore $G$ is a quasi-inverse to $F$, and $F$ and $G$ form an equivalence of categories.
[/guided]
[/step]
Explore Further
Existence of Free Groups by Reduced Words
algebra
Lie's Theorem
algebra
Semidirect Product Lie Algebra
algebra
Bounded Linear Maps Are Continuous
algebra
Biproducts Are Products and Coproducts
algebra
Quotient of a Solvable Lie Algebra Is Solvable
algebra
Existence and Uniqueness of Finite Biproducts in Additive Categories
algebra
Fully Faithful Right Adjoint Criterion
algebra