MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top


If $F$ is a left adjoint between $C$ and $D$, and $D$ has a model structure; We can define cofibrations and equivalences in $C$ to be those that are so after applying $F$. What are criterions for this to define a model structure? Where can I find a discussion? Also, the dual question about a right adjoint.

Thank you, Sasha

share|cite|improve this question
For the first question (less standard, cofibration and weak equivalence) you can check section 1.2 theorem 1.13 of The dual version is in Hope it will help you – Ilias Amrani Mar 29 '12 at 8:49
up vote 9 down vote accepted

Since in practice the dual situation occurs more often than the one you stated I will dualize your question. Given categories $\bf {D}$ and $\bf {E}$ and adjoint functors $F:\bf {D}\to \bf {E}$ and $G:\bf {E}\to \bf {D}$, with $F$ left adjoint to $G$, if $D$ is a model category then define weak equivalences and fibrations in $E$ as follows. Declare an arrow $f$ in $E$ a weak equivalence (resp. a fibration) if $G(f)$ is a weak equivalence (resp. fibration).

Then it is guaranteed that with these fibrations and weak equivalences $E$ is a cofibrantly generated model category provided the following conditions hold:

1) $D$ is cofibrantly generated. 2) The left adjoint $F$ preserves small objects. 3) The weak equivalences in $E$ contain any sequential colimit of pushouts of images $F(g)$, where $g$ is allowed to vary over the generating trivial cofibrations in $D$.

Of course condition 3 is difficult to check so it is good to know that it is satisfied automatically if:

$E$ has a functorial fibrant replacement and $E$ has functorial path objects for fibrant objects.

First appearance in the literature of such transfer principles seems to be S. E. Crans – Quillen closed model structures for sheaves, J. Pure Appl. Alg. 101 (1995), 35–57.

The more recent (and where I learned of the answer) article "Axiomatic homotopy theory for operads", by Berger and Moerdijk can be found here: where the result is used to establish a model structure on (one-object) operads.

Of course a dual answer will provide a transfer principle for fibrantly generated model categories. However, as interesting/important fibrantly generated model categories are very rare such a transfer result is hardly useful (showing that mathematics is not self-dual).

share|cite|improve this answer
Thank you! So if I assume that my model categories come with functorial factorizations (triv cofib, fib) , (cofib, triv fib) I am OK with (3). What about (2), is there some easy general widely satisfied criterion for that? – Sasha Mar 29 '12 at 16:54
Ah, I am sorry. You mean that $E$, which we still do not know to be a model category, should be shown to have functorial factorization, in order not to check (3). – Sasha Mar 29 '12 at 18:56
One useful criterion for (2) is that the right adjoint $G$ preserves (sufficiently highly) filtered colimits. – Mike Shulman Mar 29 '12 at 18:58
Downvote me if I am wrong, but it seems that I don't need $E$ to have "functorial path objects for fibrant objects" as written above, but just "path objects for fibrant objects" (all the rest the same). – Sasha Apr 1 '12 at 16:28

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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