Co-induced actions for topological and filtered groups
Jacques Darn\'e (LPP, IF)

TL;DR
This paper demonstrates that certain categories of topological and filtered groups are locally algebraically cartesian closed by establishing the existence of co-induced actions and functors, enhancing the understanding of their categorical structure.
Contribution
It proves that the category of strongly central series admits co-induced actions and shows the existence of co-induction functors in topological groups, establishing new categorical properties.
Findings
Category of strongly central series is LACC
Existence of co-induction functors in topological groups
A convenient category of topological groups is LACC
Abstract
In this note, we show that the category of strongly central series admits co-induced actions, which means that it is Locally Algebraically Cartesian Closed. We also show that some co-induction functors exist in the category of topological groups, and that a convenient category of topological groups is LACC.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsHomotopy and Cohomology in Algebraic Topology · Advanced Topics in Algebra · Algebraic structures and combinatorial models
Co-induced actions for topological and filtered groups
Jacques Darné
(December 1, 2018)
Abstract
In this note, we show that the category introduced in [Dar18] admits co-induced actions, which means that it is Locally Algebraically Cartesian Closed [Gra12, BG12]. We also show that some co-induction functors exist in the category of topological groups, and that a convenient category of topological groups is LACC.
Introduction
The present paper deals with the study of actions in some categories of topological and filtered groups. The author’s main motivation is the study of strongly central filtrations on groups (also called -series). These occur in several contexts. For instance:
- •
On the Torelli subgroup of the automorphisms of free groups, there are two such filtrations defined in a canonical way ; the Andreadakis problem, still very much opened, asks the question of the difference between them.
- •
On the Torelli subgroup of the mapping class groups of a surface, the Johnson filtration is a -series. The difference between this filtration and the lower central series is linked to invariants of -manifolds, such as the Casson invariant.
- •
On the pure braid groups or the pure welded braid groups, such filtrations appear in the study of Milnor invariants and Vassiliev invariants.
Considering strongly central filtrations as a category has led to a better understanding of phenomena appearing in these various contexts, in particular the role of Johnson morphisms, or semi-direct product decompositions of certain associated Lie algebras. At the heart of this work lies the study of actions in this category. Our main result here is a further step in that direction:
Theorem 2.1.
The category of strongly central filtrations admits co-induction functors along any morphism , that is, a right adjoint to the restriction of -actions along , for all .
Moreover, our methods can be adapted to the case of topological groups, to show:
Theorem 3.1.
In the category of topological groups, there are co-induction functors along any morphism between locally compact groups.
If we restrict our attention to a convenient category of topological spaces, then more is true:
Theorem 3.2.
A convenient category of topological groups admits co-induction functors along all morphisms.
This last result can also be seen as an enriched version of the existence of co-induction in the category of groups, obtained from Kan extensions.
Strongly central filtrations and actions
Strongly central filtrations
We recall the definition of the category introduced in [Dar18].
Definition 0.1**.**
A strongly central filtration is a nested sequence of groups such that for all . These filtrations are the objects of a category , where morphisms from to are those group morphisms from to sending each into .
Recall from [Dar18] that this category is complete, cocomplete, and homological, but not semi-abelian. It is also action-representative.
Actions
In a protomodular category , an action of an object on an object is a given isomorphism between and the kernel of a split epimorphism with a given section {Y}$${B} . Objects of endowed with a -action form a category, where morphisms are the obvious ones. This category of -objects is the same as the category of split epimorphisms onto with given sections, also called points over , and is denoted by .
For example, an action in the category of groups is a group action of a group on another, by automorphism (see for instance [BJK05]). The equivalence between -groups (in the last sense) and -points is given by:
[TABLE]
The same construction allows us to identify the category of -points in topological groups with the category of topological groups endowed with a topological -action, i.e. a group action such that is continuous.
Recall from [Dar18, Prop. 1.20] that in , an action of on is the data of a group action of on such that for all , where the commutator is taken in the semi-direct product , that is: .
Restriction and induction functors
Suppose that our protomodular category admits finite limits and colimits. Let be a morphism in . We can restrict a -action along by pulling back (in ) the corresponding epimorphism. This defines a restriction functor (also called base-change functor):
[TABLE]
If we are given a -action, we can define the induced -action by pushing out (in ) the corresponding section. This defines an induction functor , which is left adjoint to .
A question then arises naturally: is there a co-induction functor ? That is, when does the restriction functor also have a right adjoint ? When all have this property, the category is called Locally Algebraically Cartesian Closed (LACC). This is a rather strong condition, implying for instance algebraic coherence [CGV15, Th. 4.5]. This condition has been studied for example in [Gra12, BG12].
Acknowledgements: The author thanks Tim Van der Linden for asking the question which became the starting point of this work, and for inviting him in Louvain-la-Neuve, where we had the most interesting and helpful discussions. He also thanks Alan Cigoli for taking part in these discussions and making some very interesting remarks.
Contents
1 Co-induction in the category of groups
Our aim is to show that the category is LACC, that is, that it admits co-induced actions. We first review the case of groups, that will be the starting point of our construction.
The construction of co-induced group actions follows easily from the following fact: if is a group, the category of -groups identifies with the functor category , where is considered as a category with one object. Then the restriction functor along a group morphism is given by precomposition by (interpreted as a functor between the corresponding small categories). Since is complete and co-complete, this functor has both a left and a right adjoint, given by left and right Kan extensions along .
Let us describe the right adjoint of in this context. If is a functor between small categories, recall that the right Kan extension of along is given by the end formula:
[TABLE]
where denote the co-tensor over : if is a set, and , then is the product of copies of .
In our situation (, and is a -group), the coend is taken over the one object of . Moreover, is the group of applications from to (whose product is defined pointwise), and the coend is the subgroup of applications satisfying for all and , that is, the subgroup of -equivariant applications . The action of on is given by . Thus, we recover:
Proposition 1.1**.**
[Gra12, Th. 6.11]. In the category of groups, co-induction along a morphism is given by where is the group of -equivariant applications from to (with multiplication defined pointwise), on which acts by .
2 Co-induction in the category
The construction described above in the category of groups seems to be very specific, as it relies on an identification between the categories of points and functor categories. Such an identification does not hold in the case of strongly central filtrations. However, we will be able to compare the categories of points with functor categories. This will allow us to use Kan extensions for constructing co-induction.
2.1 Categories of points and functor categories
When dealing with a strongly central filtration , we will often omit the subscript , for short, denoting the underlying group by .
Let be a strongly central filtration. There is an obvious forgetful functor, recalling only that acts by automorphisms preserving the filtration:
[TABLE]
Since the compatibility conditions it forgets are only conditions on objects, not on morphisms, this functor is fully faithful. Let be a morphism in . The restriction functor between the corresponding functor categories fits into a commutative diagram:
[TABLE]
Moreover, since is complete, has a right adjoint, given by the right Kan extension . It follows from the description of limits in [Dar18, Prop. 1.10] and from the construction of Section 1 that is , that is, the filtration defined pointwise on (which is obviouly strongly central).
If we construct a right adjoint to the forgetful functor (for any ), then by composing adjunctions, will be right adjoint to , and will be right adjoint to , because is fully faithful:
[TABLE]
We will construct this functor in the next section. This will finish the proof of our first main theorem:
Theorem 2.1**.**
There are co-induction functors in . Explicitly, co-induction along is given by:
[TABLE]
which is the largest strongly central filtration smaller than on which acts.
2.2 Maximal -filtration
Since the forgetful functor is fully faithful, if it admits a right adjoint , then the counit has to be a monomorphism. Thus, if is a strongly central filtration on which acts by automorphisms preserving the filtration, we need to construct the maximum sub-object of such that the restricted action of satisfies the conditions defining a -action. We will do so through a limit process, restricting to smaller and smaller subgroups endowed with smaller and smaller filtrations.
Proposition 2.2**.**
Let and be strongly central filtrations, and let act on by group automorphisms preserving the filtration . Then there exists a greatest one among those strongly central filtrations such that the action of on induces and action of on .
Corollary 2.3**.**
The forgetful functor has a left adjoint , the filtration being the greatest filtration constructed in the above proposition.
Proof.
Suppose that acts on , that the group acts on (by filtration-preserving automorphisms), and that is a -equivariant morphism. Then acts on . As a consequence, , so that is in fact a morphism from to . ∎
Lemma 2.4**.**
Under the hypothesis of the proposition, the filtration defined by:
[TABLE]
is strongly central, and stable under the action of .
Proof.
For short, denote by . These are subgroups of because of the normality of and the formula . The three subgroups lemma applied in tells us that is contained in the normal closure of and . These are both inside , which is normal in because it is normal in and -stable. Thus , which means exactly that . Each is also stable under the action of , since if , using that the and the are -stable, we have that , thus . ∎
Proof of Proposition 2.2.
Using Lemma 2.4, define a descending series of strongly central filtrations by:
[TABLE]
Then define to be their intersection. We claim that is the requested filtration. Firstly, acts on it ; if , then:
[TABLE]
so that by taking the intersection on ,
[TABLE]
Secondly, if acts on , then by we show that for all : by definition, , and if for some , then:
[TABLE]
proving that , and our claim. ∎
3 Co-induction for topological groups
Our argument for strongly central filtrations can be adapted to the case of topological groups. The first part, about Kan extension, is exactly the same. The comparison between the category of points and functors categories, however, does only work with some restrictions : we need either the acting group to be locally compact, or the category of topological spaces to be a convenient one, for example the category of compactly generated weakly Hausdorff spaces. The main idea behind the construction is the same as before: restricting to smaller and smaller subgroups, and refining more and more the topology.
3.1 Categories of points and functor categories
A topological group is the data of a group and a topology on it, with the usual compatibility requirements [Bou71, Chap. 3] ; it can be seen as a group object in the category of topological spaces. We will denote such an object by , or only by or by , whenever the rest of the data is clear from the context. Also, in any topological group and any point , we will denote by the set of neigbourhoods of in . Note that the only topologies we will consider are the ones compatible with group structures ; expressions such as ”the finest topology satisfying…” have to be understood with this requirement in mind. Topological groups, together with continuous morphisms, form a category , which is complete, cocomplete, and homological.
If is a topological group, the obvious forgetful functor from to is fully faithful. If is a continuous morphism, then restriction along is defined in the usual way, between the category of points and between the categories of functors. The picture is exactly the same as in the case of strongly central filtrations, and the right Kan extension along is defined in the same way: , endowed with the topology inherited from the product topology on . We will show below (Proposition 3.3) that the forgetful functor admits a right adjoint for any locally compact . Thus we will have proved:
Theorem 3.1**.**
There are co-induction functors along morphisms between locally compact groups in . Explicitly, co-induction along is given by:
[TABLE]
which is the largest topological group endowed with an monomorphism into , on which acts continuously.
Note that it is not the largest topological subgroup, as largest here also mean: endowed with the coarsest possible topology.
If we restrict our category of topological spaces (for example to compactly generated weakly Hausdorff spaces), then we will show (Proposition 3.9) that our construction works for all group object .
Theorem 3.2**.**
There are co-induction functors in the category of topological groups (that is, it is LACC) when the base category of topological spaces is the category of compactly generated weakly Hausdorff spaces, or any convenient category of topological spaces, is a sense made precise below.
3.2 Maximal -topology
Suppose that is locally compact (note that we do not imply that it is locally Hausdorff). We now construct the right adjoint to the forgetful functor, in a series of lemmas leading to the proof of:
Proposition 3.3**.**
For any locally compact topological group , the forgetful functor from to has a right adjoint.
We will make use of the classical :
Proposition 3.4**.**
Let be a locally compact topological space. Then has a right adjoint, given by , where is the set of continuous maps from to , endowed with the compact-open topology.
Idea of proof.
It is easy to see that a map is continuous iff takes values inside , and is continuous. The ”only if” part crucially uses the fact that is locally compact. ∎
Consider an element of . This means that is a topological group endowed with an action of the discrete group via continuous automorphisms. If is a subgroup, consider the map:
[TABLE]
obtained by restriction of the action of on . How can we endow with a new topology such that is continuous ? In order to do this, we first need to restrict to a subgroup of . Indeed, if is continuous (for any topology on ), then for any , the map has to be continuous. Thus we are led to the definition:
[TABLE]
Lemma 3.5**.**
The subset is a subgroup of , stable under the action of .
Proof.
Consider , the set map from to adjoint to the action of on . Since is a topological group, the subset of continuous maps is a subgroup of (both are endowed with the pointwise law). Moreover, since is a topological group, it is -stable under the -action . Since acts on by group automorphism, is a group morphism. Moreover, it is also -equivariant, as one easily checks. Thus is a -stable subgroup of . ∎
We want to define the coarsest topology on making continuous. Such a topology exists, as it is the coarsest topology making continuous, which is exactly the subspace topology on . We denote it by . Remark that is finer than the subgroup topology on , because is exactly the inclusion , that has to be continuous when is endowed with .
Lemma 3.6**.**
The group acts on , whence also on , by continuous automorphisms.
Proof.
The action of on is given by pre-composition by , which is continuous. By functoriality of , it is continuous. Moreover, since the group law is defined pointwise on the target, it acts via an automorphism:
∎
Lemma 3.7**.**
Let act topologically on a topological group , and be a continuous -equivariant map. Then and is continuous.
Proof.
The map is the composite , so it is continuous. Its adjoint is also continuous and, since it coincides with , it factorizes as a set map through , which means that takes values in . Moreover, is a continuous map from to the subspace of . ∎
Proof of Proposition 3.3.
Let be a topological group on which acts by continuous automorphisms. We iterate the construction described above. This is possible thanks to Lemmas 3.5 and 3.6. We denote by the -th iterate , and we define as the intersection of the , endowed with the reunion of the . That is, is the (topological) projective limit of the .
We first show that the action of on is topological, that is, that is continuous. This is equivalent to the map being continuous for every . But this last map can be seen as the composite , and these maps are continuous by construction, so the action of on is indeed topological.
Now suppose that acts topologically on , and is a continuous -equivariant map. Then, thanks to Lemma 3.7, is again a continuous -equivariant map, and by iterating the construction, we see that is, for all . Thus takes values in and is continuous with respect to . Since any continuous -equivariant map comes uniquely from such an (which is continuous because the injection of into is, since it is ), we have showed that is the right adjoint we were looking for. ∎
3.3 Restricting to a convenient category of spaces
If is any topological group acting on another topological group , there does not seem to be a coarsest topology on making continuous, so our construction fails to produce an adjoint to the corresponding forgetful functor. However, we can change that by restricting to a convenient category of topological spaces. We need this category to satisfy the following four hypotheses:
- •
is a full subcategory of topological spaces containing the one-point space.
- •
admits (small) limits.
- •
is cartesian closed.
- •
If a subset of an object is given, there should be a topology on such that is in , the injection is continuous, and every in such that defines a continuous map . Such a topology is called the -subspace topology on .
Fact 3.8**.**
[Str09, Prop. 2.12, Lem. 2.28 and Prop. 2.30]. These hypotheses are satisfied if is the category of compactly generated Hausdorff spaces.
The point, being final in , is the unit of the cartesian monoidal structure. Thus the forgetful functor to sets is . In particular, if we denote by the right adjoint to , the underlying set of is the set of continuous maps from to :
[TABLE]
Moreover, the underlying set of a limit is the limit of the underlying diagram to sets. Indeed, if is a small category and is a diagram, then :
[TABLE]
The reader can check that the constructions of the previous paragraph work well under these hypotheses, replacing the category by the category of group objects in . Precisely, the topology has to be the -subspace topology on , and has to be the limit of the in . Thus we can state:
Proposition 3.9**.**
For any topological group , the forgetful functor from to has a right adjoint.
Remark that this proposition, together with the following fact, suggest that is a nice category to work with.
Fact 3.10**.**
The category is also action-representative: a representant of actions on is the set of continuous automorphisms , endowed with the -subspace topology.
A remark on -denriched categories
Theorem 3.2 can be obtained directly, in a fashion similar to the construction of co-induction for groups. To do that, we use the language of -enriched categories [Kel05]. The category is -enriched ; moreover, every -group can be considered as a -category with one object, and there is an obvious equivalence:
[TABLE]
Thus the same construction as in ordinary groups (Section 1) works here, replacing Kan extensions by enriched Kan extensions. This uses the fact that is -complete (that is, it is complete and co-tensored over ).
This gives an alternative proof of Theorem 3.2. However, neither Theorem 3.1 nor Theorem 2.1 fits in this machinery. Moreover, Proposition 3.9 is still meaningful in this context: it provides a right adjoint to the forgetful functor from the category of enriched functors to the category of non-enriched ones.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[BG 12] Dominique Bourn and James R. A. Gray. Aspects of algebraic exponentiation. Bull. Belg. Math. Soc. - Simon Stevin , 19(5):821–844, 2012.
- 2[BJK 05] Francis Borceux, George Janelidze, and Gregory M. Kelly. Internal object actions. Commentat. Math. Univ. Carol. , 46(2):235–255, 2005.
- 3[Bou 71] Nicolas Bourbaki. Éléments de mathématique. Topologie générale I. Chap. 1 à 4. Nouvelle edition. Paris: Hermann. XV, 351 p. F 96.00 (1971)., 1971.
- 4[CGV 15] Alan S. Cigoli, James R. A. Gray, and Tim Van der Linden. Algebraically coherent categories. Theory Appl. Categ. , 30:1864–1905, 2015.
- 5[Dar 18] Jacques Darné. On the stable Andreadakis problem. ar Xiv:1711.05991 , 2018.
- 6[Gra 12] James R. A. Gray. Algebraic exponentiation in general categories. Appl. Categ. Struct. , 20(6):543–567, 2012.
- 7[Kel 05] Gregory M. Kelly. Basic concepts of enriched category theory. Repr. Theory Appl. Categ. , (10):vi+137, 2005. Reprint of the 1982 original [Cambridge Univ. Press, Cambridge; MR 0651714].
- 8[Str 09] Neil P. Strickland. The category of CGWH spaces. https://neil-strickland.staff.shef.ac.uk/courses/homotopy/cgwh.pdf , 2009.
