![MathJax Logo](/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif)
ABSTRACT
In this paper we study grouplike monoids, defined as being monoids that contain a group to which we add an ordered set of idempotents. We classify finite categories with two objects having grouplike endomorphism monoids, by presenting a construction theorem for such categories, and proving that every grouplike category comes from this construction. Studying the algebraic properties of the endomorphism monoids allows us to gather extra information on the category itself, which in particular helps the counting problem because the nature of the monoids affects greatly the structure of the category. At the end of the paper, we give a count of certain categories with grouplike monoids, concluded from the properties of grouplike monoids that are studied in the paper.
1. Introduction
Finite categories are categories with a finite set of objects and a finite set of morphisms. The main purpose of our work in general (Ghannoum, Citation2022) thesis is the classification of finite categories. One associates to a finite category a square matrix such that the entries of the matrix are the numbers of morphisms between each pair of objects.
A more informative description of a finite category is obtained using the endomorphism monoids. We classify monoids of certain orders, then look at which categories have a given set of endomorphism monoids. For example, two monoids are called connected if there exists a category with two objects and non-empty morphism sets in both directions, where the monoids
and
are the endomorphism monoids of the two objects. Each object in the category is going to correspond to one of the monoids of endomorphisms that we specify. A groupoid is a category in which every morphism is invertible.
Starting from this idea, we can define matrices associated to categories in a different way. Instead of putting the number of morphisms between objects, we represent them in terms of the endomorphism monoids and bimodules of the category. For example a category with two objects
will be represented as follows:
where
such that A and B are monoids with a specific algebraic structure.
In this perspective the off-diagonal parts of the matrices (L and R above) are seen as bimodules over the corresponding endomorphism monoids, taking only the multiplication tables of and
. In some cases, the number of bimodules obtained specifies in a unique way the number of categories that could be obtained by two bimodules (we give some examples in section 5).
We study in this paper a specific type of finite categories called grouplike categories. To define grouplike categories, we need to define grouplike monoids first. A grouplike monoid is a monoid of the form where G is a group and
such that
We denote it by . A grouplike category is a category whose endomorphism monoids are grouplike.
Counting associative structures has been a problem for years. We introduce some of the previous work in this area:
In 2009, A. Distler and T. Kelsey counted the monoids of orders eight, nine and ten. They weren’t able to achieve more than that as the number of semigroups of order 10 was unknown (Distler & Kelsey, Citation2009).
In 2012, the number of semigroups of order 10 was known by A. Distler, C. Jefferson, T. Kelsey, and L. Kotthoff (Distler et al., Citation2012).
In 2014, G. Cruttwell and R. Leblanc introduced the question: How many categories are there with n morphisms? It means with a total number of morphisms distributed between objects in an arbitrary way. They compared the numbers obtained with the number of monoids of order n, which lead to almost the same numbers up to order 10.
In 2017, S. Allouch and C. Simpson counted the categories whose number of morphisms between each pair of objects is 2. They were able to get an exact count up to order 3, and bounds for a general size (Allouch & Simpson, Citation2018).
This work is inspired by our results with W. Fussner, T. Jakl and C. Simpson in (Fussner et al., Citation2020). Using the program Mace4 (McCune, Citation2003), we gave a count of the categories associated to the matrix
We classified monoids of order 3 and presented the number of categories between each pair of monoids. The data obtained from the count showed that the nature of the endomorphism monoids affects the classification in a very important way, which led us to prove some properties about certain monoids in a category, and to propose classifying finite categories in terms of their endomorphism monoids. In particular, in this paper we study groups and monoids that contain a group. The classification of such monoids gives very interesting properties inside the category and gives us lots of additional information allowing us to compute the number of categories with such monoids in some cases.
We present a construction theorem for grouplike categories, and we prove that every grouplike category comes from this construction. We prove that having grouplike monoids as endomorphism sets affects the structure of the bimodules, and in this way, the category is almost determined with a strong part of its properties being known.
The notion of grouplike monoids can be generalized in the following way: instead of adding k identities to a group G, we add k different groups consecutively, leading to what could be then called an iterative grouplike monoid. This idea of generalization was suggested by Sam van Gool during a seminar at IRIF in Paris. Although we don’t pursue that here, it will furnish the opportunity for future work in the direction of the results of the present paper.
Using the notation for grouplike monoids defined above, we now describe in more detail the results. The first definition encapsulates the condition that will appear in Theorem 4.11.
Definition 1.1.
Let L be a -bimodule. We say that L is i-unigen if
and as a left module and
as a right module.
Remark 1.2.
The definition of Li is given in (EquationEquation L(L)
(L) ), and the definition of
is given in Definition 2.6 and Definition 2.7.
Lemma 1.3.
Suppose L is an i-unigen -bimodule with Li as above, then there exists an isomorphism
such that
as
-bimodule. If i > 0 then the isomorphism is unique. If i = 0 then the isomorphism is well defined up to inner automorphisms.
Because of this lemma we can assume that . In this case we say that a
-bimodule L is strongly i-unigen if it is i-unigen and the isomorphism of Lemma 1.3 can be taken as the identity.
Remark 1.4.
The proof of Lemma 1.3 is on page 19.
Definition 1.5.
Let
be the algebraic matrix (Definition 2.3) of a grouplike category , such that
and
. We define
to be the maximum index i such that there exist
such that
and
.
Let be a category with one object
and endomorphism set
. Let
be two elements, and using the constant function
we obtain a new category that could be denoted
, with
and
.
If we apply this construction to a grouplike monoid , we get a category associated to the matrix
of similar copies of . Such a category will be called groupoid-like.
Theorem 1.6.
Let G be a group and a grouplike monoid of the form
such that
. Let
be a matrix of the form
of similar copies of , let
be the groupoid-like category associated to this matrix. Then we can extend the endomorphism sets
and we obtain a category
associated to the matrix
such that for all , we have
. Same for
. And there exists
a finite category associated to
.
Now let , suppose that we have the matrix
such that L and R are strongly i-unigen bimodules. Then
And (EquationEquation M(3)($M^{(3)}$)
($M^{(3)}$) ) is a matrix of a unique grouplike category, denote it by
, such that
. For all
, we have
.
We have
Theorem 1.7.
Every grouplike category comes from the construction described in Theorem 1.6.
Remark 1.8.
The proofs of Theorems 1.6 and 1.7 are on pages 19 and 20.
1.1. Organization of the paper
This paper is organized as follows:
In section 2, we give the definitions of grouplike monoids, bimodules and grouplike categories.
Since a grouplike monoid has two parts: a group G and a set I of idempotents, then we divide the discussion of each part into two sections:
In section 3, we discuss the group part of the category and how the action of the group on the set of morphisms affects the category its structure.
In section 4, we discuss the set of idempotents part which is formed of n elements, and we give the maximum element
such that a grouplike category can be determined. This leads to the proofs of Theorems 1.6 and 1.7.
In section 5, we give some examples and we apply the results in the previous sections in order to obtain a count of the number of certain grouplike categories.
2. Preliminaries
In this section we introduce grouplike categories. These are groupoids (groups in the case of monoids) in which we add extra elements, having some specific properties, to their set of morphisms. The goal is to study the structure of these categories in order to make the classification problem easier and clearer.
Definition 2.1.
A monoid A is a set equipped with a binary operation such that · is associative and there exists an identity element e such that for every element
, the equations
and
hold.
Definition 2.2.
A bimodule is a set with actions on the left and the right of the respective monoids, such that the actions commute i.e. . It can be seen as a category such that one of the sets of morphisms is empty, it’s called an upper triangulated category.
Definition 2.3.
(Carboni et al., Citation1987; Koslowski, Citation1997; Leinster, Citation1999a, Citation1999b, Citation2002). Let be a finite category. Then we get a matrix where the diagonal entries are monoids and the off diagonals entries are bimodules. This matrix is called the algebraic matrix.
Definition 2.4.
A semicategory is a category without identity morphisms.
Lemma 2.5.
(Allouch & Simpson, Citation2014) Let be a semicategory, let
with
, then we can add a morphism uʹ to
such that
and uʹ duplicates u for all composition operations. Then we get a new semicategory
with
and
.
On the other hand, we can also add the missing identities in to obtain a category
.
The previous lemma shows that we can add morphisms consecutively to a category and obtain a new category (provided that we add identities too). In our case, we define a category whose objects have grouplike endomorphism monoids, we only add elements consecutively to the monoids. The elements are idempotents and identities to the elements of the monoids. This means that each time we add an identity element, the previous identity is no longer an identity.
We need the notion of bimodules in order to make the classification and counting problem easier.
Definition 2.6.
Let be a semigroup. Whether or not
has a multiplicative identity element, we let e be a fresh element and
Then is a semigroup if the multiplication of
is extended by stipulating
for all
. More generally, if
is a semigroup, we recursively define semigroups
for
by:
If is a group, we say that the semigroups
,
, are grouplike.
Definition 2.7.
We say that a category is called a grouplike category with groups Gi if its endomorphism monoids are grouplike monoids of the form .
Definition 2.8.
A band S is an idempotent semigroup, i.e. for all ,
.
A semilattice is a commutative band.
Remark 2.9.
The set of idempotents in a grouplike monoid along with the group identity form a semilattice.
Proof.
Let be the set of idempotents where e0 is a group identity. I is a semilattice:
commutative: for
,
.
idempotent:
.
3. Group action and the orbits of the sets of morphisms
Let be a category with n objects
. For each object
there exists a monoid
and a monoid
and two operations
and
which represent the left monoid action of Ai and the right monoid action of Aj on the set of morphisms from Xi to Xj. Note here that these are the entries of the matrix defined in Definition 2.3.
In our work here, we take monoids of the form , specifically grouplike, which means that each monoid contains a subgroup. This subgroup does not act on the whole set of morphisms from Xi to Xj, but it acts on a subset of the previous set (it’s important to note here that when we say group action we mean that the identity condition holds). In the following, we introduce how the group action works, and what are exactly the subsets that the group acts on. We will be considering categories with two objects.
To avoid large paragraphs every time we have a grouplike category, we give a notation in which we specify its exact form and the structure of its grouplike monoids.
Notation 3.1.
We denote by the matrix
of monoids and bimodules such that and
, where G1 and G2 are groups and
and
such that the elements of I1 and I2 satisfy (EquationEquation (Ord)
(Ord)
(Ord) ). We denote by e0 and f0 the identities of the groups G1 and G2. Let
be the set of grouplike categories associated to M whose objects are are
such that the monoids and the bimodules are not empty. If
then
,
,
and
.
Remark 3.2.
More precisely if M is an algebraic matrix i.e. a matrix of monoids and bimodules, then Cat(M) is the set of pairs where
is a category and β is an isomorphism between the algebraic matrix of
and M. We usually don’t include this notation of β in our discussion.
Remark 3.3.
Denote by ni the order of the group Gi. When we write and
, this means that the groups G and Gʹ have the same order n.
Lemma 3.4.
Let (Notation 3.1) be a grouplike category. Then:
G1 acts on
and
.
G2 acts on
and
.
Proof.
Let
For all
and
we have
.
For all
and
we have
.
The same holds for the other sets.
Moreover, the orbit of the set is itself. Indeed, the orbit of
is a subset
. It remains to prove the other direction. Let
, then
Similarly to the case where we have groups as objects, we can conclude that the group action on these sets is free.
Proposition 3.5.
Let (Notation 3.1) be a grouplike category. The actions of the group G1 on
and
and the group G2 on
and
are all free.
Proof.
Let and
. Suppose
. If we multiply both sides by
, we obtain:
where has 2 possibilities:
If
where
, then it’s sufficient to multiply by g−1 on both sides to prove that
.
If
where
, then if we multiply by
on both sides, we get
.
Corollary 3.6.
Let (Notation 3.1) be a grouplike category. The cardinal of each orbit in the set of morphisms is equal to the order of the group acting.
Proof.
Use Proposition 3.5.
This corollary could greatly help the enumeration problem here, because it tells us the number of possibilities in some blocks of a category, such as and
for example, then we can compute how many times the multiplication of morphisms appear to obtain non isomorphic copies of blocks.
Lemma 3.7.
Let (Notation 3.1) be a grouplike category. For all
there exists at least one
such that
and vice versa.
Proof.
Let and
. Suppose
:
If
, then
If
is equal to an idempotent e in
then
Now since we have two group actions on the sets of morphisms, we want to understand the relation between these actions over the morphisms sets.
Lemma 3.8.
Let (Notation 3.1) be a grouplike category. Then
and
for all .
Proof.
Let ,
Same for the second inequality.
Lemma 3.9.
Let (Notation 3.1) be a grouplike category. Let
, the orbit of x by the action of G1 is the same orbit of x by the action of G2, i.e.
.
Proof.
Suppose . From Lemma 3.4, there is a group action by G1 on
, and by Corollary 3.6
Similarly, there is a group action by G2 on , then again by Corollary 3.6 we have
but
Therefore, , but we have
, then we obtain that n1 should be equal to n2 and
and
, hence
Proposition 3.10.
Let (Notation 3.1) be a grouplike category. For all
we have
Proof.
Let ,
and since , hence equality.
Proposition 3.11.
Let (Notation 3.1) be a grouplike category. Then the multiplication of the elements in the orbit of L and the elements in the orbit of R is the group G, i.e.
and
Proof.
: evident.
: Let
where
and
.
Definition 3.12.
Let be a category with objects
. Suppose that
is never equal to an identity for all
, then we can always reduce the category
to a new semi-category
such that
and
.
Similarly, we can eliminate morphisms that are not identities and still obtain a new semi-category.
The reason we want to eliminate some morphisms is because when we take a category whose objects have grouplike endomorphism monoids, then we could restrict the category to a smaller category with only groups as objects and the orbits in the sets of morphisms. Following this technique leads to proving some matrix properties about the coefficients on the off-diagonals. It will also clarify how such categories are built.
From the above lemmas and propositions, we conclude that we can divide each set of morphisms into two sets: the orbit of the set and the other elements that are not inside the orbit.
3.1. Groupoids
Definition 3.13.
In category theory, a groupoid generalizes the notion of group in several equivalent ways. A groupoid can be seen as a:
group with a partial function replacing the binary operation;
category in which every morphism is invertible. A groupoid with only one object is a usual group.
In the previous sections, we have discussed the structure of a grouplike category with two objects. From this category, we can extract a subcategory, which is exactly a groupoid, and we prove the following:
Theorem 3.14.
In every grouplike category with two objects, there is a sub-semicategory that is a groupoid with two objects, whose groups are the groups of the grouplike monoids.
Proof.
Let be a grouplike category with objects
. Let
and
.
the sub-semicategory of
is of the form:
Identities of
are 1G and
Let
,
with
associative
Corollary 3.15.
Let (Notation 3.1) be a grouplike category. The category
determines an isomorphism between G and Gʹ. The isomorphism is well defined up to inner automorphisms.
Proof.
Evident.
4. The sets of idempotents
In this section we study the role of the idempotent elements in the monoids, the idea is to interpret their action on the sets of morphisms.
Let (Notation 3.1) be a grouplike category. Let:
and
be the sets of idempotents of and
where
and
. And let:
and
be the sets of morphisms that are fixed by ei and fj.
Notation 4.1.
and
.
In general we have
Lemma 4.2.
and
.
Proof.
We always have
then
Now let then
and
hence
.
In the following lemma, we present some properties of the multiplication of the sets of morphisms by idempotent elements.
Lemma 4.3.
Let (Notation 3.1) be a grouplike category, we will denote by
elements of L and by
elements of R.
and
for all
,
and
.
and
for all
and
.
and
for all
or
.
If
then
and vice versa.
If
then
. (This result is also proved in Lemma 4.4).
If
and
then
and
.
In this case we can always assume that
and
(because x and y could be replaced with
and
respectively).
Proof.
Let
and
, we have:
Same for
.
for all
then
.
Let
such that
then
then
.
Suppose that
then
then
(because
) giving a contradiction.
Let
and
such that
. Suppose that
. Then
but
and
, it means that
, hence we get a contradiction. Therefore,
.
From part (4), we can see that if
then
. Suppose that
and
. Let us prove that
. We have
By the free action of G2 on
, we obtain that
.
thus
thus
.
In addition, by part (b)
and
, then we can assume that
.
From Lemma 4.3 (4) (5), we see that if one of the multiplications is an idempotent then the other way around should be an idempotent as well (we mean that it is not in the group part). That’s why in the following, we study the structure of the category whenever we have two elements such that their multiplications in both directions are idempotents.
Lemma 4.4.
Let (Notation 3.1) be a grouplike category. Let i be the maximum index such that there exist
and
;
. And let jʹ be the maximum index such that there exist
and
;
. We can assume following Lemma 4.3 (6) that
,
,
and
.
If i = 0 then
and vice versa.
If
then
and
. In addition,
and
.
Proof.
For part (1), suppose i = 0, in this case by Lemma 4.3 (4) we have is an idempotent, and by maximality of i it has to be e0. Therefore
Therefore, .
For part (2), suppose , then from Lemma 4.3 (4), suppose that
and
. From our assumption following (Lemma 4.3 (6)), we get
and
By maximality we have and
. We have
We obtain that
Then
Therefore since
, then there exists
(by maximality of i) such that
. Then
.
Which implies that . Hence
and
.
Similarly we can prove that and
.
Corollary 4.5.
In the case of Lemma 4.4 (2), we have and
.
Proof.
.
Where . Then
.
Similarly we prove that .
Remark 4.6.
As a conclusion of Lemma 4.4 and Corollary 4.5, There is a maximum element i and a maximum element j such that there exist and
such that
and
.
Proposition 4.7.
Let (Notation 3.1) be a grouplike category. Let i and j be the maximum elements such that there exist
and
;
and
. Then
and
Proof.
We want to prove that .
We always have that and
. We prove the other direction.
Let
Similarly we prove the others.
Theorem 4.8.
Let (Notation 3.1) be a grouplike category.
Suppose that i and j are the maximum elements (in the sense of Lemma 4.4) such that and
. By Lemma 4.3 (6) we can assume that
Then we can construct a sub-semicategory of the form
where and
, such that
is the maximum sub-semicategory of this form.
Proof.
is a category:
– Objects: X and Y
– Morphisms:
and
– Composition:
* Let
and
then we can write
, then
* Let
and
then
and
, then
– Identities: ei and fj
is the maximum category of this form:
If we have two elements xʹ and yʹ outside of Lij and Rji such that
and
then by Lemma 4.4 we have
and
. This is in contradiction with the maximality of i and j. Hence
is the maximum such category.
Corollary 4.9.
The monoids and
are isomorphic. We call
a groupoid-like category.
Proof.
There exist such that
and
the identities. It follows that i = j and the groups G1 and G2 are isomorphic.
Proposition 4.10.
Let
be a matrix of a grouplike category . By Theorem 4.8, we can construct a sub-semicategory
associated to the matrix
of monoids and bimodules. For all and all
we have
Proof.
Suppose that then there exists
such that
. This is in contradiction with the maximality of i.
Conclusion: Let
be a matrix of a grouplike category whose objects X and Y. Then
.
There exists
such that there exist
and
.
There exists
such that there exist
and
.
, and we obtain the following matrix
of a sub-semicategory
of
whose objects X and Y are isomorphic. Thus
and
The bimodule L has the property
i.e. L is i-unigen.
and the bimodule R has the property
i.e. R is i-unigen. Where
.
For all
,
.
The isomorphisms in 4 are inverses.
The multiplications of the elements of L by the elements of R are determined once we fix x and y.
Theorem 4.11.
Let be two groups,
, L, R be
-bimodules, and i;
,
such that the following collection EquationEquation P(i)
(P(i))
(P(i)) of properties holds:
and such that there is such that xi determines an isomorphism
and an isomorphism
Similarly for , determining an isomorphism
.
The isomorphisms and
are assumed to be inverses for the property EquationEquation P(i)
(P(i))
(P(i)) .
Then we get a category with algebraic matrix
such that ,
and
(L and R are i-unigen).
For , then the choice of xi is unique and hence the category is unique.
For i = 0, then the choice of x0 is not unique but the category is unique once x0 and y0 are fixed.
Proof.
The multiplications and
are given by the bimodule structure of L. Similarly for R.
Let ,
,
Similarly for . We can check that the multiplication is associative.
If we fix the matrix
and we take , then for the cases
For i = 0, the choices of x0 and y0 are not unique, see Remark 4.13.
Remark 4.12.
The condition EquationEquation P(i)(P(i))
(P(i)) is what we call i-unigen in Definition 1.1 (Thanks to Carlos Simpson for suggesting the terminology i-unigen).
Remark 4.13.
When i = 0, once we fix x0, there are maybe several choices for y0 that lead to inverse isomorphisms. The set of choices of y0 is given by the center of the group. The cardinal of the set of pairs in
with
is equal to the cardinal of the center of the group.
Proof of Lemma 1.3
Let L be an i-unigen bimodule. Let such that
, where
. Then
:
Suppose ,
in
. Then since
, we get
for all
, then
. Then
but they have the same cardinal, then
We define an isomorphism
as follows
If and
then
;
. Set
Then
By uniqueness of h, we have , and Lemma 1.3 is proved.
Proof of Theorem 1.6
Using Theorem 4.11 we prove Theorem 1.6, as we have the same construction of a category. In Theorem 4.11 we start by choosing to get to the algebraic matrix (EquationEquation M(3)
($M^{(3)}$)
($M^{(3)}$) ).
For the multiplication table of , the multiplication of
on L and R is given by the bimodule structure. It remains to find the maps
Let and
,
where and
are in Li and Ri and these compositions are given by
. And as in the conclusion part (7), we get the uniqueness of the category.
Proof of Theorem 1.7
From Theorem 4.8 and Proposition 4.10 we can prove Theorem 1.7. As in Remark 4.9 we prove that i and j should be the same, then the algebraic matrix obtained is
where Li and Ri are isomorphic to (conclusion part (3)). Then M is the matrix (EquationEquation M(1)
($M^{(1)}$)
($M^{(1)}$) ).
We conclude that if we have a grouplike category then the two bimodules L and R are i-unigen and the resulting isomorphisms between these two bimodules are inverses. If we then identify the groups via these isomorphisms, we can say that L and R become strictly i-unigen. Then we get the structure described in Theorem 1.6, and Theorem 1.7 is proved.
Remark 4.14.
The condition that the isomorphisms should be inverses in Theorem 4.11 is very important to obtain the grouplike category. We give a counter example to show its importance (example 4.16), but first we define the category of a matrix that is a combination of two categories associated to matrices of bimodules.
Notation 4.15.
Let
be a matrix of bimodule L, and let be a category associated to
.
Similarly let
be a matrix of bimodule R, and let be a category associated to
.
We denote by the matrix of the form
of monoids and bimodules L and R in and
. We denote by
a category associated to N.
Example 4.16.
Consider the matrix
where
such that
then is the matrix of the bimodule L. Let
be a bimodule category associated to
with the above table of multiplication.
Consider the matrix
where
such that
and
Similarly, is a matrix of the bimodule R. Notice that we changed the multiplication table of
by the involution of the group
. Let
be a bimodule category associated to
with the above table of multiplication.
But the matrix
with the above bimodules and
doesn’t admit a grouplike category. This was shown by calculating using Mace4. It also follows from Conclusion part (6) since the isomorphisms given by L and R are not inverses.
We can conclude now the general structure of grouplike categories with only 2 objects .
5. Applications
Remark 5.1.
The number of monoids is known until order 10 (Jipsen, CitationOEIS, peterjipsen, CitationOEIS). It is not easy to get a count of higher order using Mace4 as it is limited to very small orders. Some AI and machine learning techniques are now applied to classify some semigroups as in (Simpson, Citation2021). In our applications below, we need the list of monoids of order 3 as we are going to give a count of some categories having 3 morphisms in each morphism set.
List all the 7 monoids of size 3:
![](/cms/asset/227abc8a-4e36-4b9f-816a-63a3061c63c6/oama_a_2275345_uf0001_oc.jpg)
We combine monoids of three elements together in one category, two monoids are called connected if there exists a category where the monoids
and
are the endomorphism monoids of the two objects. Viewing them as objects, each object is one of the monoids of endomorphisms listed before, and this graph of a category is associated to the matrix:
From EquationEquation Table 1(1)
(1) , we can remark that
and
are grouplike monoids. Where
is of the form
such that
, and
is of the form
such that
is the trivial group.
Example 5.2.
The number of categories between and itself Let
be a category with two objects X and Y associated to the matrix
such that ,
,
and
. Fix
the number of categories associated to M1 is 64 (calculated by Mace4 [14]). But Since
is a grouplike category containing the trivial group
, then the orbit of L has size 1 and it’s unique. This means that
is unique for all
. Suppose it’s equal to 4. This reduces the number of possibilities to 15.
Similarly, let
be a category with two objects X and Y associated to the matrix
such that
,
,
and
.
For the same reason, the number of possible categories associated to M2 with the orbit condition is 15.
Then a category
associated to the matrix
such that
,
,
and
, has 15 left bimodules
and 15 right bimodules
.
We have two cases:
(a) | The bimodules are the same By Theorem 4.11, there are exactly 15 possibilities for this matrix (up to isomorphism). | ||||
(b) | The bimodules are different such that There are |
Hence for i = 0, we have 120 categories.
2. |
|
(a) | The bimodules are the same and we have 2 categories that could be associated to this matrix. | ||||
(b) | The bimodules are different and we have 1 category (up to isomorphism). |
Hence in total there are categories between
and itself.
Example 5.3.
The number of categories between ![](//:0)
and itself
The number of categories between and itself can also be found bimodules. Let
be a category with two objects X and Y associated to the matrix
such that ,
,
and
.
And similarly let be a category with two objects X and Y associated to the matrix
such that ,
,
and
.
Then there exists a category with two objects associated to the matrix
such that ,
,
and
. Where
is a left bimodule and
is a right bimodule.
Since there is 1 left and 1 right bimodule (also calculated by Mace4) and since , then the number of possible y0 is equal to the order of the center of the group
which is 2 (Theorem 4.11 and Remark 4.13). Hence there are 2 categories between
and itself.
Definition 5.4.
A category is called reduced if there do not exist two distinct isomorphic objects in
.
Lemma 5.5.
Let be a reduced category associated to the matrix
If there exists and
such that
, then
and B is a sub-monoid of A disjoint from
.
Proof.
Let
We have
ϕ is injective, indeed, suppose that , then
Remark 5.6.
We note that if then
and
is not reduced.
Proposition 5.7.
Let and
be a category associated to M whose objects are X and Y. Then
Proof.
Suppose that and
. Consider the algebraic matrix of
such that and
.
Let be a subcategory of
defined in the following way
(Objects of
): X and Y.
(Morphisms of
):
and
. (It means that we take the orbits of L and R by the action of
).
(Composition): by Proposition 3.11 we have that the composition of orbits goes into the group
.
(Identities): 1A and
.
We obtain that is associated to the following algebraic matrix
Since there exist and
such that
, then from Lemma 5.5 we have that
is a sub-monoid of A disjoint from
. Therefore,
.
Theorem 5.8.
Let be a strictly positive matrix and let
be a reduced category associated to M with
.
If there exists such that
then
.
Proof.
If is a category associated to M then every regular sub-matrix of M of size 2 is associated to a full sub-category of
[2] with two objects, then we apply Proposition 5.7 to get the result.
The classification of finite categories depends mostly on the algebraic structure of the endomorphism monoids. Monoids that are groups or contain a group impose some restrictions on the cardinality of the sets of morphisms. The properties obtained by studying such monoids make the classification and the counting problem easier. Counting finite structures is not easy in general, but optimizing the number of categories that could be obtained is a good start. We are now studying other types of monoids hoping to get classification theorems similar to the ones we have seen in this paper.
Acknowledgements
I would like to thank Wesley Fussner for he is the one that suggested working on grouplike monoids and categories, and Angel Toledo for the reference to Leinster’s paper. A very big thanks to my advisor Carlos Simpson for his constant help during my thesis.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
Supported by the Agence Nationale de la Recherche program 3ia Côte d’Azur ANR-19- P3IA-0002, and European Research Council Horizons 2020 grant 670624 (Mai Gehrke’s DuaLL project).
Additional information
Funding
References
- Allouch, S., & Simpson, C. (2014). Classification des matrices associées aux catégories finies. Cahiers de Topologie et Géométrie Différentielle Catégoriques, 55, 205–240. https://doi.org/10.1080/00927872.2017.1404081
- Allouch, S., & Simpson, C. (2018). Classification of categories with matrices of coefficient 2 and order n. Communications in Algebra, 46(7), 3079–3091. https://doi.org/10.1080/00927872.2017.1404081
- Carboni, A., Kasangian, S., & Walters, R. (1987). An axiomatics for bicategories of modules. Journal of Pure and Applied Algebra, 45(2), 127–141. https://doi.org/10.1016/0022-4049(87)90065-X
- Distler, A., Jefferson, C., Kelsey, T., & Kotthoff, L. (2012). The semigroups of order 10. Proceedings of the International Conference on Principles and Practice of Constraint Programming, Milano, Springer (pp. 883–899).
- Distler, A., & Kelsey, T. (2009). The monoids of orders eight, nine and ten. Annals of Mathematics and Artificial Intelligence, 56(1), 3–21. https://doi.org/10.1007/s10472-009-9140-y
- Fussner, W., Ghannoum, N., Jakl, T., & Simpson, C. (2020). Classification of finite semigroups and categories using computational methods. Proceedings of the 5th Conference on Artificial Intelligence and Theorem Proving AITP 2020, Aussois, France.
- Ghannoum, N. (2022, December). Investigation of finite categories [ Theses]. Université Côte d’Azur ; Université Libanaise.
- Jipsen, P. https://math.chapman.edu/jipsen/uajs/Mon.html
- Koslowski, J. (1997). Monads and interpolads in bicategories. Theory and Applications of Categories, 3(8), 182–212.
- Leinster, T. (1999a). Fc-multicategories. arXiv preprint math/9903004.
- Leinster, T. (1999b). Generalized enrichment for categories and multicategories. arXiv preprint math/9901139.
- Leinster, T. (2002). Generalized enrichment of categories. Journal of Pure and Applied Algebra, 168(2–3), 391–406. https://doi.org/10.1016/S0022-4049(01)00105-0
- McCune, W. (2003). Mace4 reference manual and guide. arXiv preprint cs/0310055.
- The on-line Encyclopedia of integer sequences. https://oeis.org/A058129.
- Simpson, C. (2021). Learning proofs for the classification of nilpotent semigroups. https://doi.org/10.48550/arXiv.2106.03015