Congruences in &SEMIGROUPS; can be described in several different ways:<P/>
<List>
<Item>
Generating pairs -- the minimal congruence which contains these pairs
</Item>
<Item>
Rees congruences -- the congruence specified by a given ideal
</Item>
<Item>
Universal congruences -- the unique congruence with only one class
</Item>
<Item>
Linked triples -- only for simple or 0-simple semigroups (see below)
</Item>
<Item>
Kernel and trace -- only for inverse semigroups
</Item>
<Item>
Word graph -- only for congruences created via
<Ref Attr="IteratorOfLeftCongruences" Label="for a semigroup"/>
or <Ref Attr="IteratorOfRightCongruences" Label="for a semigroup"/>
</Item>
<Item>
Wang pairs -- only for graph inverse semigroup
</Item>
</List>
The operation <Ref Func = "SemigroupCongruence"/> can be used to create any
of these, interpreting the arguments in a smart way. The usual way of
specifying a congruence will be by giving a set of generating pairs, but a
user with an ideal could instead create a Rees congruence or universal
congruence.<P/>
If a congruence is specified by generating pairs on a simple, 0-simple, or
inverse semigroup, then the congruence may be converted automatically to one
of the last two items in the above list, to reduce the complexity of any
calculations to be performed. The user need not manually specify, or even be
aware of, the congruence's linked triple or kernel and trace.
We can also create left congruences and right congruences, using the <Ref
Func="LeftSemigroupCongruence"/> and <Ref Func="RightSemigroupCongruence"/>
functions.<P/>
Please note that congruence objects made in &GAP; before loading the
&SEMIGROUPS; package may not behave correctly after &SEMIGROUPS; is loaded.
If &SEMIGROUPS; is loaded at the beginning of the session, or before any
congruence work is done, then the objects should behave correctly.<P/>
The main operations and attributes for congruences in the &GAP; library are:
<List>
<Item>
<Ref Attr="EquivalenceClasses" BookName="ref"/>
</Item>
<Item>
<C>NrEquivalenceClasses</C>
</Item>
<Item>
<Ref Oper="EquivalenceClassOfElement" BookName="ref"/>
</Item>
</List>
<Section>
<Heading>Congruences on Rees matrix semigroups</Heading>
This section describes the implementation of congruences of simple and 0-simple
semigroups in the &SEMIGROUPS; package, and the functions associated with
them. This code and this part of the manual were written by Michael Young.
Most of the theorems used in this chapter are from Section 3.5 of
<Cite Key = "Howie1995aa"/>.<P/>
By the Rees Theorem, any 0-simple semigroup <M>S</M> is isomorphic to a
<E>Rees 0-matrix semigroup</E> (see <Ref Chap = "Rees Matrix Semigroups"
BookName = "ref" />) over a group, with a regular sandwich matrix.
<Alt Not = "Text">
That is, <Display>S \cong \mathcal{M} ^ 0[G; I, \Lambda; P], </Display>
where <M>G</M> is a group, <M>\Lambda</M> and <M>I</M> are non-empty
sets, and <M>P</M> is regular in the sense that it has no rows or columns
consisting solely of zeroes.
</Alt>
<Alt Only = "Text">
That is, <C>S</C> is isomorphic to <C>M ^ 0[G; I, L; P]</C> where <C>G</C>
is a group, <C>L</C> and <C>I</C> are non-empty sets and the matrix
<C>P</C> is regular in the sense that it has no rows or columns
consisting solely of zeroes.
</Alt><P/>
The congruences of a Rees 0-matrix semigroup are in 1-1 correspondence with
the <E>linked triple</E>, which is a triple of the form <C>[N, S, T]</C> where:
<List>
<Item>
<C>N</C> is a normal subgroup of the underlying group <C>G</C>,
</Item>
<Item>
<C>S</C> is an equivalence relation on the columns of <C>P</C>,
</Item>
<Item>
<C>T</C> is an equivalence relation on the rows of <C>P</C>,
</Item>
</List>
satisfying the following conditions:
<List>
<Item>
a pair of <C>S</C>-related columns must contain zeroes in
precisely the same rows,
</Item>
<Item>
a pair of <C>T</C>-related rows must contain zeroes in precisely
the same columns,
</Item>
<Item>
if <C>i</C> and <C>j</C> are <C>S</C>-related, <C>k</C> and
<C>l</C> are <C>T</C>-related and the matrix entries
<M>p_{k, i}, p_{k, j}, p_{l, i}, p_{l, j} \neq 0</M>, then
<M>q_{k, l, i, j} \in N</M>, where
<Display>q_{k, l, i, j} = p_{k, i}
p_{l, i} ^ {-1} p_{l, j} p_{k, j} ^ {-1}.</Display>
</Item>
</List>
By Theorem 3.5.9 in <Cite Key = "Howie1995aa"/>, for any finite 0-simple
Rees 0-matrix semigroup, there is a bijection between its non-universal
congruences and its linked triples. In this way, we can internally
represent any congruence of such a semigroup by storing its associated
linked triple instead of a set of generating pairs, and thus perform many
calculations on it more efficiently.<P/>
If a congruence is defined by a linked triple <C>(N, S, T)</C>, then a
single class of that congruence can be defined by a triple <C>(Nx, i / S, k
/ S)</C>, where <C>Nx</C> is a right coset of <C>N</C>, <C>i / S</C> is
the equivalence class of <C>i</C> in <C>S</C>, and <C>k / S</C> is the
equivalence class of <C>k</C> in <C>T</C>. Thus we can internally
represent any class of such a congruence as a triple simply consisting of a
right coset and two positive integers.<P/>
An analogous condition exists for finite simple Rees matrix semigroups
without zero.<P/>
<Section Label="Congruences on inverse semigroups">
<Heading>
Congruences on inverse semigroups
</Heading>
This section describes the implementation of congruences of inverse
semigroups in the &SEMIGROUPS; package, and the functions associated with
them. This code and this part of the manual were written by Michael
Young. Most of the theorems used in this chapter are from Section 5.3 of
<Cite Key = "Howie1995aa"/>.<P/>
The congruences of an inverse semigroup are in 1-1 correspondence with its
<E>congruence pairs</E>. A congruence pair is a pair <C>(N, t)</C> such
that:
<List>
<Item>
<C>N</C> is a normal subsemigroup of <C>S</C> -- that is, a
self-conjugate subsemigroup which contains all the idempotents of
<C>S</C>,
</Item>
<Item>
<C>t</C> is a normal congruence on <C>E</C>, the subsemigroup of all
idempotents in <C>S</C> -- that is, a congruence on <C>E</C> such that
if <M>(e, f)</M> is a pair in <C>t</C>, then the pair
<M>(a ^ {-1} e a, a ^ {-1} f a)</M> is also in <C>t</C>,
</Item>
</List>
satisfying the following conditions:
<List>
<Item>
If <M>ae \in N</M> and <M>(e, a ^ {-1} a) \in t</M>, then <M>a \in N</M>,
</Item>
<Item>
If <M>a \in N</M>, then <M>(aa ^ {-1} , a ^ {-1} a) \in t</M>.
</Item>
</List>
By Theorem 5.3.3 in <Cite Key = "Howie1995aa"/>, for any inverse semigroup,
there is a bijection between its congruences and its congruence pairs. In
this way, we can internally represent any congruence of such a semigroup by
storing its associated congruence pair instead of a set of generating
pairs, and thus perform many calculations on it more efficiently.<P/>
If we have a congruence <C>C</C> with congruence pair <C>(N, t)</C>, it
turns out that <C>N</C> is its <E>kernel</E> (that is, the set of all
elements congruent to an idempotent) and that <C>t</C> is its <E>trace</E>
(that is, the restriction of <C>C</C> to the idempotents). Hence, we refer
to a congruence stored in this format as a "congruence by kernel and
trace".
See <C>cong_by_ker_trace_threshold</C> in Section
<Ref Sect = "Options when creating semigroups"/>
for details on when this method is used. <P/>
A Rees congruence is defined by a semigroup ideal. It is a congruence on a
semigroup <C>S</C> which has one congruence class equal to a semigroup
ideal <C>I</C> of <C>S</C>, and every other congruence class being a
singleton. <P/>
<Section>
<Heading>Universal and trivial congruences</Heading>
The linked triples of a completely 0-simple Rees 0-matrix semigroup describe
only its non-universal congruences. In any one of these, the zero element
of the semigroup is related only to itself. However, for any semigroup
<M>S</M> the universal relation <M>S<Alt Not = "Text"> \times </Alt>
<Alt Only = "Text">x</Alt>S</M> is a congruence; called the <E>universal
congruence</E>. The universal congruence on a semigroup has its own unique
representation.<P/>
Since many things we want to calculate about congruences are trivial in the
case of the universal congruence, this package contains a category
specifically designed for it, <C>IsUniversalSemigroupCongruence</C>. We also
define <C>IsUniversalSemigroupCongruenceClass</C>, which represents the single
congruence class of the universal congruence.<P/>
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung ist noch experimentell.
