I just have a question on product of 2 objects in category theory.

Given a category $\mathcal{C}$ we would like to define a product of 2 objects $A,B \in \mathcal{C}$. The definition says that $D \in \mathcal {C}$ is a product of $A$ and $B$ if for all $E\in \mathcal{C}$ there exists a unique morphisme $m:C\to D$ such that $p\circ m=f$ and $q\circ m=g$ with $p:D\to A$, $q:D\to B$, $f:E\to A$ and $g:E\to B$.

I would like to understand why this definition allows us for real to say that $D$ is a product of $A$ and $B$. I understand that given the triplet $(p,q,D)$ is not enough to say that $D$ is a "good" product; so we need to add some conditions on projections and $D$. Those conditions are exactly the unicity of morphism $m$ and applications $f,g$. But, I don't really see why adding these conditions allow to define the product of 2 objects "correctly". If someone could clarify all of this and give his intuition on this, I would really appriciate it. Thank you in advance

  • $\begingroup$ What properties does the cartesian product $A \times B$ of two sets have? How about the direct product of groups? What's common is your definition. $\endgroup$
    – Randall
    Feb 4 at 13:39
  • $\begingroup$ @Randall I think I'm starting to see what's going on. If I keep in mind the cartesian product in cat $\textbf{Sets}$ and use the same notations as in the post, I need like to "redefine" the product $D$ by arrows and so for this I need an auxiliary object $E$ which has as well applications to $A$ and $B$ because I know that $D$ is a "composition" of elements of $A$ and $B$. Is it correct? But now, why the morphism $E\to D$ must be unique? And so, if I reformulate, why the product might be necessarily a terminal object ? $\endgroup$
    – Daniil
    Feb 4 at 13:52
  • 1
    $\begingroup$ By putting "with $p:D\to A, q:D\to B, f:E\to A$ and $g:E\to B$" all at the end of your "definition" of product, you've lumped $p,q,f,g$ together in a way that confuses the logical structure of the definition. You should go back to the actual definition and understand that $p$ and $q$ are, along with $D$, part of the structure of the product, while $f$ and $g$ are, along with $E$, universally quantified in the definition. $\endgroup$ Feb 4 at 16:51

1 Answer 1


Here is the short answer: The conditions which you describe make sure that $\mathcal C(z,a\times b) \to \mathcal C(z,a) \times \mathcal C(z,b)$, $f \mapsto (\pi_af,\pi_bf)$ is an isomorphism for all $z$. This is why you can say sentences like: A map into $a\times b$ is the same as a pair of maps (one into $a$ and one into $b$).

Here is the long version. I assume that you know about the Yoneda lemma. A lot of interesting definitions in category theory are actually representations of $\mathtt{Set}$-valued functors. A representation of a contravariant functor $F: \mathcal C \to \mathtt{Set}$ is an object $c$ in $\mathcal C$ together with a natural isomorphism $\alpha: \mathcal C(-,c) \to F$ of functors. A representation usually tells us that the maps into $c$ have a nice description, namely $F$.

For example consider $\mathcal C = \mathtt{Set}$ and $c = \{0,1\} = \Omega$. Then the set maps from $A$ into $\Omega$ correspond naturally to the subsets of $A$, and $\mathtt{Set}(-,\Omega)$ is naturally isomorphic to the power set functor.

There are many (!) examples of representable covariant and contravariant functors in category theory. You could define the product of two objects $a$ and $b$ as follows. It should be an object $a\times b$ such that maps into $a\times b$ correspond naturally to pairs of maps into $a$ and $b$. In other words you like to have a representation $\alpha: \mathcal C(-,a\times b) \to \mathcal C(-,a) \times \mathcal C(-,b)$. This determines $a\times b$ up to unique compatible isomorphisms.

Now here is the catch. Because of Yoneda's lemma giving a natural transformation $\alpha: \mathcal C(-,c) \to F$ is the same as specifying an element $\alpha(1_c)\in Fc$, and $\alpha$ is an isomorphism if and only if $\alpha(1_c) \in Fc$ is terminal in the category of elements of $F$. Now in the case of a representation $\mathcal C(-,a\times b) = \mathcal C(-,a) \times \mathcal C(-,b)$ this means that specifying the natural isomorphism is the same as giving a pair of morphisms $(\pi_a,\pi_b) \in \mathcal C(a\times b,a)\times \mathcal C(a\times b,b)$. The condition that the transformation is an isomorphism becomes the universal property of the product, which you describe in your question.

  • $\begingroup$ Very nice answer, thank you very much! $\endgroup$
    – Daniil
    Feb 4 at 19:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.