I'm having a pretty hard time with this. I'm asked to show that, in the category of sets with exactly 17 elements, no two objects have either a direct product nor direct sum. Part of me doesn't even believe this statement—but whenever I try to come up with a direct product, I get snagged.

Let $(C, \alpha, \beta)$ be a [potential] direct product of $A$ and $B$. Fix some object, $C'$, with mappings, $\alpha'$ and $\beta'$, from $C'$ to $A$ and $B$ respectively. We need a unique $\gamma: C' \rightarrow C$ such that $\alpha \circ \gamma = \alpha'$ and $\beta \circ \gamma = \beta'$.

  • $C = A \times B$ just can't work, because $A \times B$ necessarily has more than 17 elements (in fact, no Cartesian-product-like $C$ can work, because the number of elements is fixed). What about some 17-element subset of $A \times B$? (In the category of sets, $\alpha$ and $\beta$ are injective, but not necessarily surjective). But, what if $\alpha'$ and $\beta'$ are both surjective? So, that can't work, because there's no $\gamma$ that could satisfy this (doesn't that mean that, in the general category of sets, $\alpha$ and $\beta$ have to also be surjective? If they don't "touch" every element in both $A$ and $B$, then one can just define a $\alpha'$ or $\beta'$ that touches the elements $\alpha$ or $\beta$ don't—thus making impossible a direct product.)

  • Let $\alpha(c_n) = a_n$ and $\beta(c_n) = b_n$. This contains bijective $\alpha$ and $\beta$, but all we have to do is define a $C'$ such that $\alpha'(c_1) = a_1$ and $\beta'(c_1) = b_2$.

Ok. So, $\alpha$ and $\beta$ have to be bijective. Let's try to prove this via negation: Since these mappings are necessarily bijective, they have to have an inverse. Thus, $\gamma$ must be such that $\gamma = \alpha^{-1} \circ \alpha'$ and $\gamma = \beta^{-1} \circ \beta'$. To show that we can choose an object $C'$ where $\gamma$ can't make the graph commute, just choose $\alpha'$ and $\beta'$ such that $\alpha^{-1} \circ \alpha \neq \beta^{-1} \circ \beta$.

  1. Can I assume that such an $\alpha'$ and $\beta'$ will always exist?
  2. Was there no point to the number of elements being $17$ specifically? This all seems to work for any category of sets with a fixed number of elements.
  3. Is there something crucial I'm missing?
  • $\begingroup$ What are the morphisms of your category? $\endgroup$ – J.-E. Pin Jul 21 '15 at 16:53
  • $\begingroup$ Just set-mappings; the same as the morphisms in the more general category of sets. $\endgroup$ – AmagicalFishy Jul 21 '15 at 17:01
  • 1
    $\begingroup$ The product projections in $\mathsf{Set}$ don't have to be epic. Consider $\emptyset \times X \to X$. $\endgroup$ – Martin Brandenburg Jul 22 '15 at 9:21
  • 1
    $\begingroup$ Notice $\emptyset \times X = \emptyset$. The unique map $\emptyset \to X$ is surjective if and only if $X=\emptyset$. This follows from the definitions of "surjective", "map", "empty set". It is not a convention. $\endgroup$ – Martin Brandenburg Jul 23 '15 at 13:02
  • 1
    $\begingroup$ In set theory, $A \times B$ is a set of ordered pairs. It itsself is not an ordered pair. If $A$ is empty, or $B$ is empty, then $A \times B$ is empty, too. $\endgroup$ – Martin Brandenburg Jul 23 '15 at 14:19

Let $n \geq 2$. If $A,B$ have a product $P$ in the category of sets with $n$ elements, then $\hom(A,P) \cong \hom(A,A) \times \hom(A,B)$ shows $n^n = n^n \cdot n^n$, a contradiction.

| cite | improve this answer | |
  • $\begingroup$ But this is assuming that the product in the category of sets is the same as the product in the category of sets with exactly 17 elements. Why can I make that assumption? $\endgroup$ – AmagicalFishy Jul 23 '15 at 0:31
  • $\begingroup$ No, I don't assume this. Recall that $X \times Y$, in any category, is defined by $\hom(-,X \times Y) \cong \hom(-,X) \times \hom(-,Y)$. $\endgroup$ – Martin Brandenburg Jul 23 '15 at 9:24
  • $\begingroup$ Oh! I didn't know this. Quickly looking up some notation, you're using hom functors, yes? (I haven't seen these yet—the exercise I was working on was in the first chapter of Mathematical Physics by Geroch; $X \times Y$ has been defined only in terms of the Cartesian Product between sets.) $\endgroup$ – AmagicalFishy Jul 23 '15 at 12:41
  • $\begingroup$ You don't have to know about hom-functors. If $(X \leftarrow P \rightarrow Y)$ is a product, you can directly verify that there is a bijection $\hom(T,P) \cong \hom(T,X) \times \hom(T,Y)$ for all objects $T$. (Actually this is natural in $T$, but this is irrelevant for the exercise.) $\endgroup$ – Martin Brandenburg Jul 23 '15 at 13:00
  • $\begingroup$ Oh. $hom(T,P)$ just refers to the "set" of morphisms from $T$ to $P$? What is $hom(T, X) \times hom(T, Y)$ defined categorically? (I know it's something like the Cartesian Product set-wise, but do all categories have an equivalent?) $\endgroup$ – AmagicalFishy Jul 23 '15 at 15:01

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.