nLab
solid functor

Contents

Idea

A solid functor (also called a semi-topological functor) is a forgetful functor U:AX for which the structure of an A-object can be universally lifted along sinks. One can also say that U has not just a left adjoint but all possible “relative” left adjoints.

When U is solid, colimits in A can be constructed in a natural way out of colimits in X, and A inherits strong cocompleteness properties from X.

Definition

Let U:AX be a faithful functor. A U-structured sink is a sink in X of the form (Ua if ix). Note that the indexing family iI need not be a set, it can be a proper class. A semi-final lift of such a U-structured sink consists of a morphism xgUb in X such that

  1. Every composite gf i:Ua iUb is in the image of U, i.e. is of the form U(g˜) for some g˜:a ib (necessarily unique, since U is faithful), and

  2. g is initial with this property, i.e. for any other morphism xgUb such that each gf i is in the image of U, there exists a unique h:bb in A such that g=U(h)g.

Finally, U is called solid if every U-structured sink has a semi-final lift.

<!-- Created with SVG-edit - http://svg-edit.googlecode.com/ --> Layer 1 a i a_i a j a_j U ( a j ) U(a_j) U ( a i ) U(a_i) f i f_i f j f_j b b h h g g' b b' g g U ( b ) U(b) U ( b ) U(b') U ( h ) U(h) X X

Examples

Properties

Left adjoint

For any xX, the empty family of morphisms into x is a U-structured sink, and a semi-final lift for this family is a universal arrow xUb. Therefore, if U is solid, then it has a left adjoint.

Lifting of colimits

Suppose that U:AX is solid and let F:DA be a diagram such that UF has a colimit in X, consisting of a cocone UFd ic. Let cUe be a semi-final lift of this U-structured sink, for which we have induced morphisms Fd ie in A. Since U is faithful, these morphisms are a cocone under F, and the semi-finality makes it into a colimit in A.

Therefore, if A is solid over X, then it admits all colimits which X does. Moreover, if we understand colimits in X, and we understand the semi-final lifts, then we understand colimits in A.

In particular, if X is cocomplete, then so is A. In fact, more is true: if X is total, then so is A.

Lifting of model structures

The standard transfer theorem for model structures states that if U:AX is a functor such that

  1. U has a left adjoint F,
  2. U is accessible, i.e. preserves κ-filtered colimits for sufficiently large κ,
  3. X has a cofibrantly generated model structure, and
  4. Transfinite composites of pushouts of images under F of generating acyclic cofibrations in X become weak equivalences after applying U (the acyclicity condition),

then A has a cofibrantly generated model structure in which the weak equivalences and fibrations are created by U. Using an argument of (Nikolaus) we can show:

Theorem

Let U:AX be an accessible solid functor, and assume that X has a cofibrantly generated model structure and the following acyclicity condition:

  • If F:DA is a filtered diagram and U(F(d i))x is a cocone under UF, each of whose legs is an acyclic cofibration in X, then the semifinal lift xU(b) of this U-structured sink is also an acyclic cofibration.

Then the transfer theorem applies, so that A has a cofibrantly generated model structure in which the weak equivalences and fibrations are created by U.

Proof

We have remarked above that U has a left adjoint, and we assumed it to be accessible, so it remains to show that the given acyclicity condition implies the standard one.

We first show that pushouts in A of images under F of generating acyclic cofibrations become acyclic cofibrations (not just weak equivalences) upon applying U. Let i:xy be a generating acyclic cofibration, and

F(x) f a F(i) F(y)\array{F(x) & \overset{f}{\to} & a\\ ^{F (i)}\downarrow \\ F(y)}

a diagram in A of which we would like to take the pushout. Consider the pushout of the corresponding diagram in X:

x f¯ U(a) i g y h U(a) xy.\array{x & \overset{\bar{f}}{\to} & U(a)\\ ^{i}\downarrow && \downarrow^g\\ y& \underset{h}{\to} & U(a) \sqcup_{x} y.}

Since X is a model category, g is an acyclic cofibration. Therefore, if U(a) xykU(b) is a semifinal lift of the singleton sink {g}, by assumption, k is also an acyclic cofibration and thus so is the composite U(a)U(b). But it is straightforward to verify that in fact, the map ab of which this is the image (which exists by assumption) gives a pushout diagram in A:

F(x) f a F(i) F(y) h¯ b.\array{F(x) & \overset{f}{\to} & a\\ ^{F(i)}\downarrow && \downarrow\\ F(y) & \underset{\bar{h}}{\to} & b.}

If U is not just accessible but finitary, then it preserves all transfinite composites, so any transfinite composite of such pushouts in A maps to a transfinite composite in X, and we know that transfinite composites of acyclic cofibrations in X are acyclic cofibrations, so the desired acyclicity condition follows. In general, we can argue as follows: given a transfinite sequence a 0a 1 in A of such pushouts, its colimit (= composition) can be constructed as above by forgetting down to X, taking the colimit there, and then taking the semifinal lift. But since acyclic cofibrations in X are closed under transfinite composites, the legs of the colimiting cocone in X are acyclic cofibrations. Hence by the assumed acyclicity condition, so is the semifinal lift, and hence (by composition) so are the images in X of the legs of the colimiting cone in A. This completes the proof.

References

The example of algebraic fibrant objects and the argument entering the above lifting theorem appears in

See also model structure on algebraic fibrant objects.

Revised on December 12, 2011 21:05:03 by Mike Shulman (71.136.231.40)