6
$\begingroup$

Suppose you have a pair of orthogonal factorization systems, $(E_0, M_0), (E_1, M_1)$ in a category $\cal C$ such that $M_0\subseteq M_1$; this entails that there is a ternary factorization $$ X\xrightarrow{e_1} A\xrightarrow{e_0 m_1} B\xrightarrow{m_0} Y $$ where each arrow is labeled according to the class it belongs to.

Suppose now to have another OFS $(L,R)$ on $\cal C$, and to factor the middle arrow $A\to B$ into $A\xrightarrow{l} S\xrightarrow{r} B$; this gives a factorization $X\to S\to B$.

Does this define a third factorization system $(E_{01}\wr L, M_{01}\wr R)$?

$\endgroup$
3
  • 1
    $\begingroup$ What is the meaning of the tilte symbol. Is it the set of compositions ? Is $E_{01}$ the intersection of $E_0$ and $E_1$ ? Actually it is equal to $E_1$. $\endgroup$ Apr 21, 2016 at 15:27
  • $\begingroup$ In what application does this arise? $\endgroup$ Apr 21, 2016 at 16:12
  • $\begingroup$ @PhilippeGaucher $E_{01}$ is not the intersection $E_0\cap E_1$; it is only a shorthand to denote the ternary factorization $\Delta[1]\to FS({\cal C})$ with a single symbol. $\endgroup$
    – fosco
    Apr 21, 2016 at 21:33

1 Answer 1

8
$\begingroup$

Yes.

First note that Galois connections are far more common than field theorists would have you believe: any binary relation gives rise to one and the two fixed sets a closed under any appropriate algebraic structure.

Write

  • $P\perp Q$ for "every member of $P$ is orthogonal (in the sense of factorisation systems) to every member of $Q$";

  • $P;Q$ for the set of all composites of members of $P$ followed by members of $Q$;

  • $L'=L\cap E_0\cap M_1$ and $R'=R\cap E_0\cap M_1$ (to avoid any ambiguity in this argument); and

  • $E=E_1;L'$ and $M=R';M_0$.

Then

  • $E_1\perp M_1\supset R'$,

  • $E_1\perp M_1\supset M_0$,

  • $L'\subset L\perp R\supset R'$ and

  • $L'\subset E_0\perp M_0$.

Since orthogonality respects composition, it follows that $E\perp M$, whilst by construction $E;M$ is the entire hom-class of the category.

Therefore $(E,M)$ is a factorisation system.

The things that I have assumed are all proved in my book but I don't have a copy to hand to look up the references.

In the question as stated, it is given that $M_0\subset M_1$, but there is no order-relationship between these and $R$. This is why it was necessary to introduce $R'$. The construction gives a way of handling expressions in the lattice of factorisation systems on a category, so it would be an interesting exercise in lattice theory to find out whether this is modular or even distributive.

$\endgroup$
4
  • $\begingroup$ I'm not able to see why you define $L', R'$. I also actually realized that the generality I need is $E_1\subset L\subset E_0$, I think that in this case there are simpler proofs. Anyways, your general argument is neat and elegant. Thanks! If you spend two more words on the definition of $L', R'$ I'll be happy to accept the answer. $\endgroup$
    – fosco
    Apr 21, 2016 at 21:37
  • $\begingroup$ (a rapid check after: I'm convinced by your argument; the same proof of yours seems valid in case $R\subset M_1$ ,using $(L,R)$ instead of $(L', R')$, simply because in that case $r,l\in E_0\cap M_1$) $\endgroup$
    – fosco
    Apr 21, 2016 at 21:53
  • $\begingroup$ Ah, now I see the edit; nice. I think I'll try to do the exercise, do you have an answer for it? $\endgroup$
    – fosco
    Apr 22, 2016 at 12:11
  • $\begingroup$ I am a child indeed (-: $\endgroup$
    – fosco
    Apr 22, 2016 at 12:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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