Unlike notes where one usually incorporates structure and some level of self-containment, the following write-up could be seen more like a diary and a means to guide oneโs learning journey towards some more sophisticated references. In the cases where good references are lacking, we will aim to provide full detail or accumulate material that was hard to find (or stuff that is never spelled out in other references). In that way, we hope this will be of value to someone in the future.
Mathematical objects very often come in families. One quick example is the following: Let denote the category of abelian groups, and fix a commutative ring . Then the there is a well defined category
of -modules (see also module object). Now every homomorphism of commutative rings induces two functors
which are referred to as restriction of scalars and extension of scalars, respectively. It turns out that both these assignments are functorial: We have two functors
It is often more convenient to encode the functoriality of the (covariant or contravariant) construction in a different way. In fact, to every functor , one can associate a new category , referred to as the category of elements, the Grothendieck construction, or the unstraightening? of the functor , which comes endowed with a โbundle mapโ
More generally, for an -category there is an (adjoint) equivalence of -categories
Here by โniceโ bundle maps we really mean cocartesian fibrations (in particular, there is another equivalence which encodes contravariant functors with domain and target ). This equivalence is referred to as the Straightening/Unstraightening equivalence and it is certainly one of the most foundational and important tools (maybe second only to the yoneda lemma) used to define functors in higher category theory. The idea is that, usually when we want to define a functor with values in -categories we would have to specify an infinite amount of coherences which might be completely unfeasible; thus instead we go unstraightened and find us a nice bundle map (i.e. cococartesian fibration?) that encodes the functor we are after. Some applications for this fibrational approach are e.g. the theory of stacksthe definition of monoidal -categories (or more generally for -operads), operad algebras, Covering spaces, the calculation of -(co)limits, โฆ
We shall discover that when one seeks to pursue algebra in a homotopy coherent manner, -operads present the ideal framework. In fact, all of higher algebra is written in the language of (infinity,1)-operads. But the thing that keeps the machine of -operads well oiled, or makes riding it even possible is the theory of cocartesian fibrations and in turn therefore the Straightening/Unstraightening Equivalence.
Given a functor of -categories, a natural construction is to look at the fibers of , that is, for an object, the fiber of at , denoted by is given by the pullback diagram:
Now the question is: When do the fibers have contra- or covariant dependence with regards to the category . It is certainly false that every functor induces a โstraightenedโ functor:
that retains all the information that was present in the original functor . In fact, an easy counter example is the following.
Let be the category with three objects such that there exists precisely one arrow from to and one from to . On the other hand, let be the category with two objects with precisely one arrow from to (i.e. ). Define the functor by the assignment
Put graphically, we have
where the colors indicate where objects and morphisms are mapped to. Now the fibers and are given by the terminal category and the disjoint union of two terminal categories , respectively. There is now two (different) choices for a functor
which is really just a functor . Thus the choices are either or , and both these functors are far from ever being able to recover all the information of the original functor .
We shall first give slick, quick and abstract model-independent? definitions, and then after we shall unravel these definitions (in the -categorical setting) to give some intuition.
Aaron Mazel-Gee, 2015 Let be a functor of -categories, and let be a morhpism in .
Let be a functor of -categories, and denote by (resp. ) the full subcategory of the arrow category spanned by the -cocartesian morphisms (resp. cartesian morphisms).
where the functors denote the source projections, respectively.
where the functors denote the target projections, respectively.
Let us unravel both these definitions. For to be a -cocartesian morphism really boils down to saying that for each object we have a pullback square:
Let us suppose now that are 1-categories, then the above equivalence of morphism spaces is really just a bijection of hom-sets. In particular, for this morphism to be a bijection is equivalent to the following: For every , and for every such that , there exists a unique lift of , i.e.
such that
Writing this down by means of a diagram yields:
The above graphical representation thus encodes what it means for a morphism to be -cocartesian (for a functor between -categories. Staying in the -categorical setting, let us look at the second condition, namely that we have an equivalence of categories
Here the important bit is really only that the dotted arrow is essentially surjective: The objects of are really just arrows such that there exists an so that . Hence essential surjectivity tells us that for any arrow , there exists a -cocartesian lift
of .
Let be a functor of -categories. Then a morphism is * -cocartesian if and only if * -cartesian if and only if
There is another reformulation for to be a -(co)cartesian morphism: is -cocartesian if and only if the induced functor
is a (surjective) equivalence (see Cartesian morphism Proposition 2.4).
Now assume , a functor of -categories, is a cocartesian fibration.
We define the pseudofunctor with values in the -category of small categories as follows:
which on objects assigns the fibers , while for a morphism , the functor
is defined as follows:
This yields a family of lifts which we fix. Now, define the value of the functor on the object to be the target of the lift associated to and :
implying (and thus functoriality).
Next up, we want to verify that the construction of is actually a pseudofunctor. In order to verify this, let us prove a quick lemma:
For composable morphisms of with -cocartesian and , we have that is -cocartesian if and only if is -cocartesian. In particular, -cocartesian morphisms are closed under composition.
For -cocartesian morphisms we have a pasting of pullback squares which by pasting law for pullbacks yields the desired claim:
is a pseudofunctor.
We have specified what does on objects and what it does on morphisms in . All that remains is to prove that this assignment is pseudo-functorial. For this, let be a composable pair of morphisms in and consider for an object the diagram:
where the (unique) dotted arrows are induced by the lifting property of the -cocartesian morphisms and , respectively (justified by the above Lemma). In fact, again by uniqueness the dotted arrows must be mutually inverse isomorphisms - these will be the components of our natural isomorphism witnessing functoriality.
Moreover, by uniqueness yet again, we have a commutative diagram
which proves that we have a natural isomorphism
An analogous argument shows the remaining axioms of pseudo-functoriality.
One can now define a category of cocartesian fibrations with codomain , denoted , whose objects are cocartesian fibrations and whose morphisms are โbundleโ maps which preserve cocartesian morphisms: Now the construction extends to a functor
with values in the category of pseudofunctors from to . It then turns out that this construction yields an equivalence of categories (see Grothendieck construction). We will spell this out for the more general case of -categories over the course of the next few chapters.
Lurieโs definition of (co)cartesian fibrations in the setting of quasicategories (or more generally simplicial sets) is the following:
(see Kerodon, Definition 5.1.1.1)
Let be a morphism of simplicial sets, and let be an edge of .
admits a solution, provided that .
admits a solution, provided that .
Let be a cocartesian fibration of -categories. To every morphism in the -category , we want to associate a functor (uniquely determined up to isomorphism by Proposition 5.2.2.8), which Lurie calls covariant transport functor?. We will skip a few steps and instead of proving existence of such a covariant transport functor for a single , we shall immediately construct a functorial assignment for . This is referred to as parametrized covariant transport? in Lurieโs Kerodon.
Definition 5.2.8.1 Let be a cocartesian fibration of simplicial sets and let be vertices of . We will say that a morphism
is given by parametrized covariant transport? if there exists a morphism
satisfying the following conditions:
where the lower horizontal map is induced by the inclusion by currying.
The restriction is given by projection onto , and the restriction is equal to .
For every edge of and every object , the composite map
is a -cocartesian edge of .
Let us unravel these conditions:
The first condition (the commutative diagram) really just says that evaluated at is equal to the constant diagram with value . In other words, plugging in into yields a lift for .
The second condition says that , interpreted as a natural transformation between -functors has domain and target as depicted:
As already hinted at in the previous remark: For every edge , the composite map
is given by covariant transport along (see also Definition 5.2.2.4).
(see Example 5.2.8.3.) Let denote the category of pointed sets, and let denote the forgetful functor . Then is a cocartesian fibration (in fact, it is a left covering map), whose fiber over an object may be identified with . For every pair of sets , the evaluation map
is given by parametrized covariant transport.
(see Proposition 5.2.8.4) Let be a cocartesian fibration of simplicial sets, and let be vertices of . Then we have the following: 1. There exists a morphism which is given by parametrized covariant transport?. 2. An arbitrary diagram is given by parametrized covariant transport if and only if it is isomorphic to (as an object of the -category ).
In order to prove this, we need a lemma
Let be a cocartesian fibration of simplicial sets, let be a simplicial set, and suppose we are given a lifting problem
Then we have: 1. The lifting problem admits a solution which is a -cocartesian lift of . 2. Let be any object of the -category . Then is isomorphic to (as an object of ) if and only if , where is yet another -cocartesian lift of which solves the above lifting problem.
This follows by means of the currying isomorphism and the fact that postcompositon with a cocartesian fibration is itself a cocartesian fibration. Therefore, the lifting problem reduces to . In this case, being a cocartesian fibration readily implies the first part, while Remark 5.1.3.8 does the second part.
This is a summary of Section 5.5 in Kerodon.
Given a locally Kan simplicial category? , there is a canonical slice and coslice construction and for an object . The main question is now does this behave with regards to the homotopy coherent nerve operation?
Before adressing this, one has yet another natural defintion: For locally Kan as above, one can define
locally Kan enriched left and right cone categories. For those we have canonical isomorphisms:
Let be a simplicially enriched category and let . We have a functor which sends objects to , and carries the cone point to the object . If and are objects of , then the induced map of simplicial sets
is simply given by the inclusion map. If is an object of , then the induced map
is equal to . This is referred to as the right cone contraction functor?.
Analogously we have a functor , which is referred as the left cone contraction functor?. We then have the following useful result:
Propositon 5.5.2.16 We have a bijection
induced by postcomposition with the right cone contraction functor?. The dual result holds too of course.
Now we realize that there is a map
carrying the cone point to the vertex . Therefore by the above bijection we obtain the slice comparison morphism
This map is a monomorphism, but it is generally not an isomorphism, nor is it a homotopy equivalence in all cases. However, in good situation it will be a homotopy equivalence.
Theorem 5.5.2.21 Let be locally Kan and let be an object of with the following property: For every morphism and every object , the morphism of simplicial sets
is a Kan fibration. Then the coslice comparison morphism is an equivalence of -categories.
This result is useful for the following reason: Later on in Chapter 5.5 Lurie defines a chain of -categories: where the first morphism is an equivalence by Proposition 5.5.6.6 (this is really a corollary of the above theorem).
We can describe the above categories informally as follows: All the above categories have the same collection of objects. For the morphisms we have quite different situations however: Morphisms in are really just (strictly) commutative triangles: Morphisms in are triangles that commute up to isomorphism in the -category : Morphisms in are diagrams with the comparison morphism not necessarily being an isomorphism. Finally, the last play in the game is not an -category but an -category, with the same objects and morphisms as , but with non-invertible natural transformations between functors.
For amazing Nlab lecture notes see An Introduction to Homological Algebra.
Amazing Nlab Lecture notes at Introduction to Stable Homotopy Theory.
For now see stable (infinity,1)-category.
Last revised on August 23, 2024 at 12:42:20. See the history of this page for a list of all contributions to it.