• Aucun résultat trouvé

The kernel of a monoid morphism

N/A
N/A
Protected

Academic year: 2022

Partager "The kernel of a monoid morphism"

Copied!
8
0
0

Texte intégral

(1)

LIAFA, CNRS and University Paris Diderot

The kernel of a monoid morphism

Jean-´Eric Pin1

1LIAFA, CNRS and University Paris Diderot

October 2009, Valencia

LIAFA, CNRS and University Paris Diderot

Outline

(1) Kernels and extensions (2) The synthesis theoem (3) The finite case

(4) Group radical and effective characterization (5) The topological approach

LIAFA, CNRS and University Paris Diderot

Basic definitions

An elementeof a semigroup isidempotentif e2=e. The set of idempotents of a semigroupSis denoted byE(S).

A semigroup isidempotentif each of its elements is idempotent (that is, ifE(S) =S). Asemilatticeis acommutative and idempotent monoid.

Avariety of finite monoidsis a class of finite monoids closed under takingsubmonoids,quotient monoids andfinite direct products.

LIAFA, CNRS and University Paris Diderot

Part I

Kernels and extensions

LIAFA, CNRS and University Paris Diderot

The kernel of a group morphism

Letπ:H→Gbe asurjectivegroup morphism.

Thekernelofπis the group

T = Ker(π) =π−1(1) andHis anextensionofGbyT.

Thesynthesis problemin finite group theory consists in constructingHgivenGandT.

LIAFA, CNRS and University Paris Diderot

The kernel of a group morphism

Letπ:H→Gbe asurjectivegroup morphism.

Thekernelofπis the group

T = Ker(π) =π−1(1) andH is anextensionofGbyT.

Thesynthesis problemin finite group theory consists in constructingHgivenGandT.

◮Is there a similar theory forsemigroups?

(2)

LIAFA, CNRS and University Paris Diderot

A specific example

A monoidM is anextension of a group by a semilatticeif there a surjective morphismπfromM onto a groupGsuch thatπ−1(1)is asemilattice.

• How tocharacterizethe extensions of a group by a semilattice?

• Is there asynthesis theoremin this case?

• In the finite case, what is thevarietygenerated by the extensions of a group by a semilattice?

LIAFA, CNRS and University Paris Diderot

The difference between semigroups and groups

Letπ:H→Gbe a surjectivegroupmorphism and letK=π−1(1). Thenπ(h1) =π(h2) iff

h1h−12 ∈K.

Ifπ:M→Gbe a surjectivemonoidmorphism and K=π−1(1), there is in general no way to decide whetherπ(m1) =π(m2), givenK.

For this reason, the notion of akernel of a monoid morphismhas to be stronger. . .

LIAFA, CNRS and University Paris Diderot

The kernel category of a morphism

LetGbe a group and letπ:M→Gbe a surjective morphism. Thekernel categoryKer(π) ofπhasGas its object set and for allg, h∈G Mor(u, v)={(u, m, v)∈G×M×G|uπ(m) =v}

Note thatMor(u, u)is a monoid equal toπ−1(1) and thatGacts naturally (on the left) onKer(π):

u v

gu gv

m

gm

LIAFA, CNRS and University Paris Diderot

A first necessary condition Proposition

Letπbe a surjective morphism from a monoidM onto a groupGsuch thatπ−1(1)is asemilattice.

Thenπ−1(1) =E(M)and theidempotentsofM commute.

Proof. Asπ−1(1)is a semilattice,π−1(1)⊆E(M).

Ifeis idempotent, thenπ(e)is idempotent and therefore is equal to1. ThusE(M)⊆π−1(1).

LIAFA, CNRS and University Paris Diderot

A second necessary condition

LetMbe a monoid withcommuting idempotents.

• It isE-unitaryif for alle, f∈E(M)and x∈M, one of the conditionsex=f or xe=f implies thatxisidempotent.

• It isE-denseif, for each x∈M, there are elementsx1andx2inM such thatx1xand xx2are idempotent.

Note that any finite monoid isE-dense, since every element has anidempotent power. But(N,+)is notE-densesince its unique idempotent is0.

LIAFA, CNRS and University Paris Diderot

A second necessary condition (2) Proposition

Letπbe a surjective morphism from a monoidM onto a groupGsuch thatπ−1(1) =E(M). Then M isE-unitary dense.

Proof. Ifex=f thenπ(e)π(x) =π(f), that is π(x) = 1. Thusx∈E(M)andM isE-unitary.

Letx∈M and letg=π(x). Letx¯be such that π(¯x) =g−1. Thenπ(¯xx) = 1 =π(x¯x). Therefore

¯

xxandx¯xareidempotent. ThusMisE–dense.

(3)

LIAFA, CNRS and University Paris Diderot

The fundamental group π

1

(M )

LetF(M)be thefree groupwith basisM. Then there is a natural injectionm→(m)fromMinto F(M). Thefundamental groupπ1(M)ofMis the group with presentation

hM|(m)(n)=(mn)for allm, n∈Mi Fact. IfM is anE-densemonoid withcommuting idempotents, thenπ1(M)is the quotient ofMby the congruence∼defined byu∼viff there exists an idempotentesuch thateu=ev.

LIAFA, CNRS and University Paris Diderot

Characterization of extensions of groups Theorem (Margolis-Pin, J. Algebra 1987)

LetM be a monoid whoseidempotents form a subsemigroup. TFCAE:

(1) there is a surjective morphismπ:M→G onto agroupGsuch thatπ−1(1) =E(M), (2) the surjective morphismπ:M→Π1(M)

satisfiesπ−1(1) =E(M), (3) M isE-unitary dense.

LIAFA, CNRS and University Paris Diderot

Part II

The synthesis theorem

LIAFA, CNRS and University Paris Diderot

Categories

Notation: uandvareobjects,x,y,p,q,p+x, p+x+yaremorphisms,p,q,x+y,y+xare loops.

u v

x y

p q

For each objectu, there is a loop0ubased onu such that, for every morphismxfromutov, 0u+x=xandx+ 0v=x.

Thelocal monoidatuis the monoid formed by the loopsbased onu.

LIAFA, CNRS and University Paris Diderot

Groups acting on a category (1)

Anactionof a groupGon a categoryC is given by a group morphism fromGinto the automorphism group ofC. We writegxfor the result of the action ofg∈Gon anobjectormorphismx. Note that for allg∈Gandp, q∈C:

• g(p+q) =gp+gq,

• g0u= 0gu.

The groupGactsfreelyonCifgx=ximplies g= 1. It actstransitivelyif the orbit of any object ofC underGisObj(C).

LIAFA, CNRS and University Paris Diderot

The monoid C

u

LetGbe a group actingfreelyandtransitivelyon a categoryC. Letube an object ofC and let

Cu={(p, g)|g∈G, p∈Mor(u, gu)}

ThenCuis amonoidunder the multiplication defined by(p, g)(q, h) = (p+gq, gh).

u gu

hu ghu

p

q p+gq gq

(4)

LIAFA, CNRS and University Paris Diderot

A property of he monoid C

u

Proposition

LetGbe a group actingfreelyandtransitivelyon a category. Then for each objectu, the monoidCuis isomorphic toC/G.

LIAFA, CNRS and University Paris Diderot

The synthesis theorem

Theorem (Margolis-Pin, J. Algebra 1987)

LetM be a monoid. The following conditions are equivalent:

(1) M is anextension of a group by a semilattice, (2) M isE-unitary densewithcommuting

idempotents,

(3) M is isomorphic toC/G, whereGis a group actingfreelyandtransitivelyon a connected, idempotent and commutativecategory.

LIAFA, CNRS and University Paris Diderot

The covering theorem

LetMandN be monoids withcommuting idempotents. A cover is a surjective morphism γ:M→N which induces anisomorphismfrom E(M)toE(N).

Theorem (Fountain, 1990)

EveryE-densemonoid with commuting idempotents has anE-unitary dense cover with commuting idempotents.

LIAFA, CNRS and University Paris Diderot

Part III The finite case

LIAFA, CNRS and University Paris Diderot

Closure properties Proposition

The class ofextensions of groups by semilatticesis closed under takingsubmonoidsanddirect product.

Proof. Letπbe a surjective morphism from a monoidMonto a groupGsuch thatπ−1(1)is a semilattice. IfN be a submonoid ofM, thenπ(N) is asubmonoidofGand hence isgroupH. ThusN is an extension ofH andπ−1(1)∩N is a

semilattice.

Direct products: easy.

LIAFA, CNRS and University Paris Diderot

Variety generated by finite extensions

LetVbe the variety generated byextensionsof groupsbysemilattices.

A monoid belongs to Viff it is aquotientof an extensionof a group by a semilattice.

M

N G

π−1(1) =E(M) γ π

The monoidN belongs toV.

(5)

LIAFA, CNRS and University Paris Diderot

Variety generated by finite extensions

LetVbe the variety generated byextensionsof groupsbysemilattices.

A monoid belongs toViff it is aquotientof an extensionof a group by a semilattice.

M

N G

π−1(1) =E(M) γ π

The monoidN belongs toV.

◮This diagram is typical of arelational morphism.

LIAFA, CNRS and University Paris Diderot

Relational morphisms

LetM andN be monoids. Arelational morphism fromM toN is a mapτ :M→ P(N)such that:

(1) τ(s)isnonemptyfor alls∈M, (2) τ(s)τ(t)⊆τ(st)for alls, t∈M, (3) 1∈τ(1).

Examples of relational morphisms include:

•Morphisms

•Inverses of surjective morphisms

•The composition of two relational morphisms

LIAFA, CNRS and University Paris Diderot

Graph of a relational morphism

The graphRofτ is a submonoid ofM×N. Let α:R→Mandβ:R→N be theprojections.

Thenαissurjectiveandτ =β◦α−1.

M N

R⊆M×N

α β

τ

α(m, n)=m τ(m)=β(α−1(m)) β(m, n)=n τ−1(n)=α(β−1(n))

LIAFA, CNRS and University Paris Diderot

An example of relational morphism

LetQbe a finite set. LetS(Q)thesymmetric group onQand letI(Q)be the monoid of allinjective partial functionsfromQtoQunder composition.

Letτ :I(Q)→S(Q)be the relational morphism defined byτ(f)={Bijections extendingf}

1 2 3 4

f 3 - 2 -

h1 3 1 2 4 h2 3 4 2 1

τ(f) ={h1, h2}

LIAFA, CNRS and University Paris Diderot

Relational morphisms Proposition

Letτ:M→N be a relational morphism. IfT is a subsemigroupofN, then

τ−1(T)={x∈M|τ(x)∩T 6=∅}

is a subsemigroup ofM.

In our example,τ−1(1)is asemilatticesince τ−1(1)={f∈I(Q)|theidentityextendsf}

={subidentitiesonQ} ≡(P(Q),∩)

LIAFA, CNRS and University Paris Diderot

Finite extensions and relational morphisms

A monoid belongs to Viff it is aquotientof an extension of a group by a semilattice.

M

N G

π−1(1)semilattice γ(π−1(1))semilattice γ π

τ

Proposition

A monoidN belongs toViff there is arelational morphismτ fromN onto agroupGsuch that τ−1(1)is asemilattice.

(6)

LIAFA, CNRS and University Paris Diderot

Finite extensions and relational morphisms (2)

Consider the canonical factorization ofτ:

N G

R⊆N×G

α β

τ

Thenαinduces aisomorphismfromβ−1(1)onto τ−1(1)since

β−1(1)={(n,1)∈R|1∈τ(n)}

τ−1(1)={n∈N |1∈τ(n)}

LIAFA, CNRS and University Paris Diderot

A non effective characterization

Theorem (Margolis-Pin, J. Algebra 1987)

LetN be a finite monoid. TFCAE (1) N belongs toV,

(2) N is aquotientof anextension of a group by a semilattice,

(3) N iscoveredby anextension of a group by a semilattice,

(4) there is arelational morphismτ fromN onto agroupGsuch thatτ−1(1)is asemilattice.

LIAFA, CNRS and University Paris Diderot

The finite covering theorem Theorem (Ash, 1987)

Everyfinitemonoid with commuting idempotents has afiniteE-unitary coverwith commuting idempotents.

Corollary

The varietyVis the variety of finite monoids with commuting idempotents.

LIAFA, CNRS and University Paris Diderot

Part IV Group radical

LIAFA, CNRS and University Paris Diderot

Group radical of a monoid

LetMbe a finite monoid. Thegroup radicalofM is the set

K(M)= \

τ:M→G

τ−1(1) where the intersection runs over the set of all relational morphisms fromM into afinite group.

LIAFA, CNRS and University Paris Diderot

Universal relational morphisms Proposition

For each finite monoidM, there exists a finite groupGand a relational morphismτ:M→G such that K(M) =τ−1(1).

Proof. There are only finitely many subsets ofM.

ThereforeK(M)=τ1−1(1)∩ · · · ∩τn−1(1)where τ1:M→G1, . . . , τn:M→Gn. Let

τ :M→G1× · · · ×Gnbe the relational

morphism defined byτ(m) =τ1(m)× · · · ×τn(m).

Thenτ−1(1) =K(M).

(7)

LIAFA, CNRS and University Paris Diderot

Another characterization of V Theorem

LetMbe a finite monoid. TFCAE:

(1) M belongs toV, (2) K(M) is asemilattice,

(3) TheidempotentsofM commuteand K(M) =E(M).

◮Is there an algorithm to computeK(M)?

LIAFA, CNRS and University Paris Diderot

Ash’s small theorem Theorem (Ash 1987)

IfM is a finite monoid withcommuting idempotents, thenK(M) =E(M).

Corollary

The varietyVis the variety of finite monoids with commuting idempotents.

LIAFA, CNRS and University Paris Diderot

Ash’s big theorem

Denote byD(M)the least submonoidT ofM closed underweak conjugation: ift∈T and a¯aa=a, thenat¯a∈T and¯ata∈T.

Theorem (Ash 1991)

For each finite monoidM, one hasK(M) =D(M).

Corollary

One can effectively computeK(M).

LIAFA, CNRS and University Paris Diderot

Part V

The topological approach

LIAFA, CNRS and University Paris Diderot

The pro-group topology

Thepro-group topologyonA[onF G(A)] is the least topology such that every morphism fromA on afinite(discrete) group is continuous.

Proposition

LetLbe a subset ofAandu∈A. Thenu∈L iff, for every morphismβfromAonto a finite groupG,β(u)∈β(L).

LIAFA, CNRS and University Paris Diderot

A topological characterization of K(M ) Theorem (Pin, J. Algebra 1991)

Letα:A→Mbe surjective morphism. Then m∈K(M)iff1∈α−1(m).

1∈α−1(m) ⇐⇒ for allβ:A→G,1∈β(α−1(m))

⇐⇒ for allτ :M→G,1∈τ(m)

⇐⇒ for allτ :M→G,m∈τ−1(1)

⇐⇒ m∈K(M)

(8)

LIAFA, CNRS and University Paris Diderot

Finitely generated subgroups of the free group Theorem (M. Hall 1950)

Every finitely generated subgroup of the free group isclosed.

Theorem (Ribes-Zalesskii 1993)

LetH1, . . . , Hnbe finitely generated subgroups of the free group. ThenH1H2· · ·Hnisclosed.

LIAFA, CNRS and University Paris Diderot

Computation of the closure of a set

Theorem (Pin-Reutenauer, 1991 (mod R.Z.))

There is a simple algorithm to compute theclosure of a givenrational subsetof the free group.

Theorem (Pin, J. Algebra 1991 (mod P.R.))

There is a simple algorithm to compute theclosure of a givenrational languageof the free monoid.

LIAFA, CNRS and University Paris Diderot

Another proof of Ash’s big theorem

Theorem (Pin, Bull. Austr. Math. 1988)

Given the simple algorithm to compute theclosure of arational language, one hasK(M) =D(M).

Therefore, Ribes-Zalesskii’s theorem givesanother proofof Ash’s big theorem.

Theorem

Given adecidablevarietyV, the variety generated byV-extensions of groups isdecidable.

Références

Documents relatifs

The problem of determining whether a given family of finite sets of natural numbers is a code was proved to be decidable in [3].. We will extend this result by giving a

In this section we investigate the intersection of two submonoids of A ∗ finitely generated by prefix sets by studying the product of two deterministic semi-flower automata

In type A, the tower of algebras ( K [M (S n )]) n∈ N possesses long sought-after properties. Indeed, it is well-known that several combinatorial Hopf algebras arise as

Then the closed ideals of L correspond exactly, via the dimension map, to polar subsets of the Dedekind-complete lattice- ordered group K 0 (L). This leads to methods which can

We now show that the monoid H + admits interesting combinatorial properties similar to those of F + , in connection with distinguished elements ∆ n defined as right lcms of atoms,

An algebraic proof of the fact that Artin braid groups are torsion free was given in [3].. The aim of this note is to observe that the proof of [3], which uses group pre- sentations

For each method, we obtain formulas that provide bounds on the total number of vector additions in F k b required to perform any of the following three computations, given a vector S

In [2] is mentioned a conjecture about centralizers in generalized braid groups of certain dth roots of central elements, and the Birman–Ko–Lee monoid is used to prove the