4. Isomorphism

4.1. A categorical view of bijections

If we think of morphisms as the “functions” between objects in a category, then it is natural to ask whether there is a categorical analog of a bijection.

A morphism \(f: A → B\) is called an isomorphism if there exists a morphism \(g: A → B\) such that \(g ∘ f=\operatorname{id}_A\) and \(f ∘ g=\operatorname{id}_B\). We write \(f^{-1}\) to denote \(g\) when it exists.


Exercise 4.1.1: Show that there exists a unique choice for \(g\) in the above definition, from which it follows that the notation \(f^{-1}\) is unambiguous.
Exercise 4.1.2: Show that every morphism from a object to an object in a category is an isomorphism.

In \(\mathbf{Set}\) a function is an isomorphism if and only if it is a bijection, so isomorphism is a valid generalization of bijection.

In \(\mathbf{Grph}\) the isomorphisms are the morphisms which are bijective on both vertices and edges. In \(\mathbf{Mon}\) the isomorphisms are the bijective morphisms. However, in \(\mathbf{Pos}\) the isomorphisms are not just the bijective morphisms. Consider for example the objects \(\mathcal{A}= \langle \{a,b,c\}, \leq \rangle\), \(\mathcal{B}= \langle \{u,v,w\}, \preccurlyeq \rangle\) in \(\mathbf{Pos}\) shown in the figure below.

[font=\large,semithick]
\draw[font=\Large,thick] (-.4,3) node {$A$};
\draw[gray] (-1.8,-.5) -- (1,-.5) -- (1,2.5) -- (-1.8, 2.5) -- (-1.8,-.5);
\draw (-.4,0) node {$a$};
\draw (-1.4,2) node {$b$};
\draw (.6,2) node {$c$};
\draw (-.9,1) node {\rotatebox[origin=c]{110}{$\leq$}};
\draw (.2,1) node {\rotatebox[origin=c]{65}{$\leq$}};
\path[->,bend left] (1.2,1) edge node [fill=white] {$f$} (2.8,1);
\draw[font=\Large,thick] (3.5,3) node {$B$};
\draw[gray] (3,-.5) -- (4,-.5) -- (4,2.5) -- (3, 2.5) -- (3,-.5);
\draw (3.5,0) node {$u$};
\draw (3.5,.5) node {\rotatebox[origin=c]{90}{$\preceq$}};
\draw (3.5,1) node {$v$};
\draw (3.5,1.5) node {\rotatebox[origin=c]{90}{$\preceq$}};
\draw (3.5,2) node {$w$};

Let \(f\colon \mathcal A\to \mathcal B\) be the morphism defined as follows: \(f(a) = u\), \(f(b) = v\), \(f(c) = w\). Then \(f\) is clearly bijective and preserves the ordering—i.e., \(x \leq y \implies f(x) {{\preccurlyeq}}f(y)\). However, the inverse map \(f^{-1}\colon \mathcal B\to \mathcal A\) is not order preserving, and \(\mathcal A\) and \(\mathcal B\) are not isomorphic posets.


Exercise 4.1.3: What are the isomorphisms of Rel?

Let \(\mathcal X= \langle X, \leq \rangle\) and \(\mathcal Y= \langle Y, \preccurlyeq \rangle\) be objects in \(\mathbf{Pos}\). Recall, the morphisms are the order-preserving maps; that is, \(\phi : \mathcal X\rightarrow \mathcal Y\) is a morphism of \(\mathbf{Pos}\) iff \(a \leq b\) implies \(\phi(a) \preccurlyeq \phi(b)\).

Let \(\mathcal L= \langle L, \vee, \wedge\rangle\) and \(\mathcal M= \langle M, \curlyvee, \curlywedge\rangle\) be objects in \(\mathbf{Lat}\). Then \(\psi: \mathcal L\rightarrow \mathcal M\) is a morphism in \(\mathbf{Lat}\) iff for all \(a, b \in L\) we have \(\psi( a \vee b ) = \psi(a) \curlyvee \psi(b)\) and \(\psi( a \wedge b ) = \psi(a) \curlywedge \psi(b)\).

It is not hard to see that a lattice \(\langle L, \vee, \wedge\rangle\) can be represented by a poset \(\langle L, \leq \rangle\) with respect to the partial order \(\leq\) defined by \((\forall x, y) (x \leq y\; \longleftrightarrow \; x \wedge y = x)\).


Exercise 4.1.4: Show that every lattice homomorphism is order-preserving. Show that not every order-preserving map is a lattice homomorphism.
Exercise 4.1.5: Find a nontrivial category with no nontrivial isomorphisms.
Exercise 4.1.6: Find a nontrivial category where every morphism is an isomorphism.

4.2. Monomorphism

We can also give a categorical generalization of injections.

A morphism \(f\colon A\to B\) is called a monomorphism if for every object \(X\) and every pair \(h,h'\colon X\to A\) of morphisms, \(f\circ h=f\circ h'\) implies \(h=h'\). When \(f\) is a monomorphism we often say “\(f\) is mono” and write \(f\colon A\rightarrowtail B\).


Exercise 4.2.1: Show that in Set the monomorphisms are precisely the injective functions.

A morphism \(h\colon X\to A\) is sometimes called a generalized element of \(A\). A morphism \(f\) is mono just in case it is injective on the generalized elements of its domain.

The monomorphisms \(f\) in \(\mathbf{Set}\) are precisely those that are injective on the global elements \(h\colon\mathbf{1}\to A\). Thus, in \(\mathbf{Set}\), to decide whether a morphism is mono, it suffices to check \(X=\mathbf{1}\).


Exercise 4.2.2: Does this condition suffice in Grph as well?

4.3. Epimorphism

The dual concept to monomorphism generalizes surjections to arbitrary categories.

A morphism \(f\colon A\to B\) is called an epimorphism if for every object \(Y\) and pair \(y_1,y_2\colon B\to Y\) of morphisms, \(y_1\circ f=y_2\circ f\) implies \(y_1=y_2\). When \(f\colon A\to B\) is an epimorphism we often say “\(f\) is epi” and write \(f\colon \twoheadrightarrow B\).


Exercise 4.3.1: Show that in Set the epimorphisms are precisely the surjective functions.
Exercise 4.3.2: Is there a single object that we can use in place of a generic \(Y\) to check that a function is epi in Set?

4.4. Isomorphism

Recall from above, a morphism \(f\colon A \to B\) in a category \(\mathcal C\) is said to be an isomorphism if there is a morphism \(g\colon B \to A\) such that \(f ∘ g = \mathrm{id}_B\) and \(g ∘ f = \mathrm{id}_A\). When this happens, we say that

  • \(g\) is an inverse for \(f\)
  • \(f\) is an inverse for \(g\)
  • \(A\) and \(B\) are isomorphic, denoted \(A \cong B\)
  • the pair \(f, g\) witnesses \(A\cong B\).

Example. In the category determined by a partially ordered set, the only isomorphisms are the identities, and in a preorder \(X\) with \(x, y \in X\) we have \(x \cong y\) iff \(x \leq y\) and \(y \leq x\).


4.5. Solutions

Exercise 4.1.1

Solution (to do).


Exercise 4.1.2

Solution (to do).


Solution to Exercise 4.1.3.

Let \(\psi: 𝐋 → 𝐌\) be a lattice homomorphism. Fix \(a, b ∈ L\) with \(a ≤ b\).

If \(a\) and \(b\) have a greatest lower bound (as they must in a lattice), then the statement \(a ≤ b\) is equivalent to \(a ∧ b = a\). Therefore,

\[ψ(a) = ψ (a ∧ b) = ψ(a) ⋏ ψ(b).\]

(The second equality holds since \(ψ\) is a lattice homomorphism.)

The equation \(ψ(a) = ψ(a) ⋏ ψ(b)\) is equivalent to \(ψ(a) ⪯ ψ(b)\). Thus, if \(ψ\) is a lattice homomorphism and \(a ≤ b\), then \(ψ(a) ⪯ ψ(b)\). That is, lattice homomorphisms are order preserving.

The converse is false. That is, an order preserving map need not be a lattice homomorphism. For example, consider the two lattices shown in the Hasse diagrams below.

To do: insert diagram here.

The map \(ϕ: \{0, 1, 2, 3\} → \{a, b, c\}\) defined by \(ϕ(0) = a\), \(ϕ(2) = b = ϕ(3)\), and \(ϕ(1) = c\) is clearly order preserving, but it is not a lattice homomorphism. For example, \(a = ϕ(0)\) \(= ϕ(2 ∧ 3)\) \(≠ ϕ(2) ∧ \phi(3)\) \(= b ∧ b = b\).


Exercise 4.1.4

Solution (to do).


Exercise 4.1.5

Solution (to do).


Exercise 4.1.6

Solution (to do).


Exercise 4.2.1

Solution (to do).


Exercise 4.2.2

Solution (to do).


Exercise 4.3.1

Solution (to do).


Exercise 4.3.2

Solution (to do).