5.4 Other monoids
5.4.2 Results
We will now consider a number of submonoids in turn, giving the classification of their congru- ences. The total number of congruences of each monoid is shown in Table 5.36.
Monoid Size Number of congruences
Sn n! 3 In P n k=0 n k n! (n−k)! 3n−1 POIn 2nn n+ 1 Mn Pnk=0 22nkCk n+ 7 PPn C2n n+ 7 Bn (2n−1)!! 32n+52 or 32n+ 13 Jn Cn 12n+72 or 12n+ 7 PBn P n k=0 2n 2k (2k−1)!! 3n+ 7 Pn B2n 3n+ 7
Table 5.36: The number of congruences on various diagram monoids. Numbers shown are correct forn≥5.
We saw in Example 1.81 a way in which partial transformations lie in the partition monoid
Pn. We will therefore start with three monoids of partial transformations which embed into
Pn as submonoids: Sn, In, andPOIn. For all the monoids below, we assumen≥3, since any
lowernhas very few elements and is rather trivial to solve.
Symmetric group Sn
The symmetric group Sn is isomorphic to the subgroup ofPn consisting of all bipartitions of
its normal subgroups (see Section 3.1.2). These normal subgroups are well known: the trivial group {id}, the alternating group An, the whole symmetric group Sn itself, and uniquely in
the case thatn= 4, the Klein 4-group K4 =h(1 2)(3 4),(1 3)(2 4)i. For n≥3 these normal
subgroups (and hence these congruences) are all distinct.
Symmetric inverse monoid In
Recall that the symmetric inverse monoidIn consists of all the partial permutations of rank up
tonunder composition. This embeds intoPn as in Example 1.81 as the submonoid consisting
of all bipartitions with trivial kernel and cokernel.
The ideals ofIn form a chain with respect to containment, and are precisely the sets
Ik ={α∈ Mn: rankα≤k},
fork∈ {0, . . . , n}, as is the case forPn andMn.
The congruences of In were classified in [Lib53], and are reformulated in the context of
IN-pairs as follows.
Theorem 5.37 ([EMRT18, Theorem 4.1]). Let In be the inverse symmetric monoid of degree
n, for n≥0. The congruences of In form a chain, and are as follows:
• the Rees congruencesRk corresponding to the idealsIk, for k∈ {0, . . . , n};
• the congruences RI,N corresponding to the IN-pairs (Ik−1, N) for k ∈ {2, . . . , n} and
N ∈ {K4,Ak,Sk} being any non-trivial normal subgroup of Sk (the group isomorphic to
a maximal subgroup ofJk).
Note thatAk andSk will be used for everyk∈ {2, . . . , n}, butK4 will only be used when
k = 4. Since A2 is trivial, we have RI1,A2 = R1. Note also that, since there is only one
bipartition in In of rank 0, we have R0 = ∆In. As an example, the congruences ofI6 are as follows: ∆I6=R0⊂R1 ⊂RI1,S2 ⊂R2 ⊂RI2,A3⊂RI2,S3 ⊂R3 ⊂RI3,K4 ⊂RI3,A4 ⊂RI3,S4 ⊂R4 ⊂RI4,A5⊂RI4,S5 ⊂R5 ⊂RI5,A6⊂RI5,S6 ⊂R6=∇I6.
Order-preserving partial permutation monoidPOIn
The monoidPOInof all order-preserving partial permutations embeds intoPnas the submonoid
consisting of all the planar bipartitions with trivial kernel and cokernel. An element ofPOIn is
defined by its domain and codomain, soPOIncontains nr
2 elements of rankr, orPn r=0 n r 2 = 2n n
Its ideals have the same description as forIn, and its congruences are all Rees. Hence the
congruences ofPOIn form the chain
∆POIn=R0⊂R1⊂R2⊂. . .⊂Rn=∇POIn. This is shown in [Fer01, Proposition 2.6].
Planar bipartition monoid PPn
The planar bipartition monoid PPn simply consists of all the planar bipartitions in Pn. It
therefore contains the Motzkin monoid, which has the additional restriction that a bipartition’s blocks have size 1 or 2. The planar bipartition monoidPPn has a number of elements equal to
the Catalan numberC2n [OEIS, A000108]. Its congruence lattice is in fact isomorphic to that
ofMn, and its congruences have the same descriptions (detailed in Theorem 5.23) [EMRT18,
§7].
Brauer monoid Bn
The Brauer monoidBn consists of all the bipartitions in Pn whose blocks all have size 2. The
Brauer monoid contains
(2n−1)!! = (2n−1)·(2n−3)· · ·5·3·1
elements in total [OEIS, A001147]. Since each block in Bn must have size 2, the monoid only
contains elements with rank equal to n (mod 2) – in particular, Bn never contains both an
element of rank 0 and an element of rank 1. For this reason, in classifying the congruences of
Bn, we must consider two different cases: one in whichnis odd, and one in whichnis even.
The odd case is by far the simpler. All bipartitions in Bn are odd in this case, so the
Rees congruences are {R1, R3, . . . , Rn}. We also have lifted congruences{δ1, λ1, ρ1} which are
defined in the same way as for Mn. Finally, via IN-pairs, we have RIk−2,Ak and RIk−2,Sk for eachk∈ {3,5, . . . , n}.
In the even case, there are many more congruences. All bipartitions are even, so the Rees congruences are{R0, R2, . . . , Rn}. We also have lifted congruences
{ζI0,∆, ζI0,LI0, ζI0,RI0, ζI2,∆, ζI2,LI0, ζI2,RI0},
which we denote, in a way similar to the Motzkin monoid, as{δ0, λ0, ρ0, δ2, λ2, ρ2}. Some more
congruences arise from IN-pairs: (I0,S2) gives rise to
δS2 =ζI0,S2,∆, λS2 =ζI0,S2,LI0, ρS2 =ζI0,S2,RI0, RI0,S2;
and (I2, K4) gives rise to
δK4 =ζI2,K4,∆, λK4 =ζI2,K4,LI0, ρK4 =ζI2,K4,RI0, RI2,K4.
Finally, we haveRIk−2,Ak andRIk−2,Sk for eachk∈ {4,6, . . . , n}.
Rn RI4,A6 R4 RI2,S4 RI2,A4 RI2,K4 ρK4 λK4 δK4 R2 ρ2 λ2 δ2 RI0,S2 ρS2 λS2 δS2 R0 ρ0 λ0 δ0 =∇Bn = ∆Bn Rn RI3,A5 R3 RI1,S3 RI1,A3 R1 λ1 ρ1 δ1 =∇Bn = ∆Bn
Figure 5.38: Congruence lattice of Bn forn ≥5 when n is odd (upper left) and even (lower
Jones monoid Jn
The Jones monoid Jn is the submonoid ofPn consisting of all planar bipartitions with blocks
of size 2. By this definition, we can see thatJn=PPn∩ Bn. Its size is given by the Catalan
number Cn [OEIS, A000108].
As with the Brauer monoid, we consider two different cases based on whethern is odd or even; however, the congruence lattices are much simpler. Ifnis odd, then the only congruences areδ1,λ1,ρ1, and the Rees congruences{R1, R3, . . . , Rn}. If, on the other hand,nis even, then
the congruence lattice is isomorphic to that of Mn/2, and its description can be obtained by
doubling each number in the description of that lattice. That is, the congruences are precisely
{δ0, δ2, λ0, λ2, ρ0, ρ2, R0, R2, . . . , Rn}.
These results are proven in [EMRT18, §9], and the lattices are shown in Figure 5.39. Rn R4 R2 λ2 R0 ρ2 λ0 δ2 ρ0 δ0 =∇Jn = ∆Jn Rn R5 R3 R1 λ1 ρ1 δ1 =∇Jn = ∆Jn
Figure 5.39: Congruence lattice ofJn forn≥4 when nis odd (left) and even (right).
Bipartition monoid Pn and partial Brauer monoid PBn
Finally, we can state the congruence lattice of the entire bipartition monoidPn. As in the case
of the Motzkin monoid, we have Rees congruences {R0, R1, . . . , Rn} and lifted congruences
{δ0, δ1, λ0, λ1, ρ0, ρ1}. The additional congruences onPn come from IN-pairs: the retractable
IN-pair (I1,S2) gives rise to congruences
and the non-retractable IN-pairs (Ik−1,Ak) and (Ik−1,Sk) give us the congruences RIk−1,Ak andRIk−1,Ak, fork∈ {3, . . . , n}. Uniquely fork= 4 we also have the IN-pair (I3, K4), yielding the congruence RI3,K4. These are all the congruences on Pn, as is proven in [EMRT18, §5].
The lattice is shown in Figure 5.40.
We should also mention the partial Brauer monoidPBn, the submonoid ofPn consisting of
all the bipartitions with blocks of size 1 or 2. It has
n X k=0 2n 2k (2k−1)!!
elements [OEIS, A066223], as shown in [Hd14, 2.1]. Again we can see that this monoid contains the Motzkin monoid; in fact, it is clear from the definitions that Mn = PPn ∩ PBn. Its
congruence lattice has the same description as that of Pn [EMRT18,§6], and is therefore also
Rn R3 RI2,S3 RI2,A3 R2 RI1,S2 ρS2 λS2 δS2 R1 ρ1 λ1 δ1 R0 ρ0 λ0 δ0 =∇Pn = ∆Pn
Chapter 6
Principal factors and counting
congruences
In Chapter 5 we classified the congruences of the Motzkin monoid and several related diagram monoids. This classification was achieved by first calculating the congruence lattices for small values ofnusing the computational techniques described in Chapters 2 and 4, and then building up theory in order to prove a classification for generaln. In this chapter we present some more results about congruences that were obtained in a similar way, by first looking for patterns in computational results, and then extending the results and attempting to prove them for larger semigroups. Thelibsemigroupslibrary and theSemigroupsandsmallsemipackages forGAPwere used to carry out the initial computations [MT+18, M+19, DM17, GAP18].
6.1
Congruences of principal factors
In this section, we will consider an interesting decomposition of a semigroup related to itsJ- classes: a semigroup’s principal factors. After defining this construction, we will consider the principal factors of the full transformation monoid Tn, and classify their congruences. After
this, we will look at the principal factors of some other, somewhat similar monoids, and classify their congruences using similar principles.