To use the **Circle** package first you need to load it as follows:

gap> LoadPackage("circle"); ----------------------------------------------------------------------------- Loading Circle 1.4.0 (Adjoint groups of finite rings) by Alexander Konovalov (http://www.cs.st-andrews.ac.uk/~alexk/) and Panagiotis Soules (psoules@math.uoa.gr). ----------------------------------------------------------------------------- true gap>

Note that if you entered examples from the previous chapter, you need to restart **GAP** before loading the **Circle** package.

Because for elements of the ring \(R\) the ordinary multiplication is already denoted by `*`

, for the implementation of the circle multiplication in the adjoint semigroup we need to wrap up ring elements as CircleObjects, for which `*`

is defined to be the circle multiplication.

`‣ CircleObject` ( x ) | ( attribute ) |

Let `x` be a ring element. Then `CircleObject(x)`

returns the corresponding circle object. If `x` lies in the family `fam`

, then `CircleObject(x)`

lies in the family `CircleFamily`

(3.1-5), corresponding to the family `fam`

.

gap> a := CircleObject( 2 ); CircleObject( 2 )

`‣ UnderlyingRingElement` ( x ) | ( attribute ) |

Returns the corresponding ring element for the circle object `x`.

gap> a := CircleObject( 2 ); CircleObject( 2 ) gap> UnderlyingRingElement( a ); 2

`‣ IsCircleObject` ( x ) | ( category ) |

`‣ IsCircleObjectCollection` ( x ) | ( category ) |

An object `x` lies in the category `IsCircleObject`

if and only if it lies in a family constructed by `CircleFamily`

(3.1-5). Since circle objects can be multiplied via `*`

with elements in their family, and we need operations `One`

and `Inverse`

to deal with groups they generate, circle objects are implemented in the category `IsMultiplicativeElementWithInverse`

. A collection of circle objects (e.g. adjoint semigroup or adjoint group) will lie in the category `IsCircleObjectCollection`

.

gap> IsCircleObject( 2 ); IsCircleObject( CircleObject( 2 ) ); false true gap> IsMultiplicativeElementWithInverse( CircleObject( 2 ) ); true gap> IsCircleObjectCollection( [ CircleObject(0), CircleObject(2) ] ); true

`‣ IsPositionalObjectOneSlotRep` ( x ) | ( representation ) |

`‣ IsDefaultCircleObject` ( x ) | ( representation ) |

To store the corresponding circle object, we need only to store the underlying ring element. Since this is quite common situation, we defined the representation `IsPositionalObjectOneSlotRep`

for a more general case. Then we defined `IsDefaultCircleObject`

as a synonym of `IsPositionalObjectOneSlotRep`

for objects in `IsCircleObject`

(3.1-3).

gap> IsPositionalObjectOneSlotRep( CircleObject( 2 ) ); true gap> IsDefaultCircleObject( CircleObject( 2 ) ); true

`‣ CircleFamily` ( fam ) | ( attribute ) |

`CircleFamily(fam)`

is a family, elements of which are in one-to-one correspondence with elements of the family `fam`, but with the circle multiplication as an infix multiplication. That is, for \(x\), \(y\) in `fam`, the product of their images in the `CircleFamily(fam)`

will be the image of \( x + y + x y \). The relation between these families is demonstrated by the following equality:

gap> FamilyObj( CircleObject ( 2 ) ) = CircleFamily( FamilyObj( 2 ) ); true

`‣ One` ( x ) | ( operation ) |

This operation returns the multiplicative neutral element for the circle object `x`. The result is the circle object corresponding to the additive neutral element of the appropriate ring.

gap> One( CircleObject( 5 ) ); CircleObject( 0 ) gap> One( CircleObject( 5 ) ) = CircleObject( Zero( 5 ) ); true gap> One( CircleObject( [ [ 1, 1 ],[ 0, 1 ] ] ) ); CircleObject( [ [ 0, 0 ], [ 0, 0 ] ] )

`‣ InverseOp` ( x ) | ( operation ) |

For a circle object `x`, returns the multiplicative inverse of `x` with respect to the circle multiplication; if such one does not exist then `fail`

is returned.

In our implementation we assume that the underlying ring is a subring of the ring with one, thus, if the circle inverse for an element \(x\) exists, than it can be computed as \(-x(1+x)^{-1}\).

gap> CircleObject( -2 )^-1; CircleObject( -2 ) gap> CircleObject( 2 )^-1; CircleObject( -2/3 ) gap> CircleObject( -2 )*CircleObject( -2 )^-1; CircleObject( 0 )

gap> m := CircleObject( [ [ 1, 1 ], [ 0, 1 ] ] ); CircleObject( [ [ 1, 1 ], [ 0, 1 ] ] ) gap> m^-1; CircleObject( [ [ -1/2, -1/4 ], [ 0, -1/2 ] ] ) gap> m * m^-1; CircleObject( [ [ 0, 0 ], [ 0, 0 ] ] ) gap> CircleObject( [ [ 0, 1 ], [ 1, 0 ] ] )^-1; fail

`‣ IsUnit` ( [R, ]x ) | ( operation ) |

Let `x` be a circle object corresponding to an element of the ring `R`. Then the operation `IsUnit`

returns `true`

, if `x` is invertible in `R` with respect to the circle multiplication, and `false`

otherwise.

gap> IsUnit( Integers, CircleObject( -2 ) ); true gap> IsUnit( Integers, CircleObject( 2 ) ); false gap> IsUnit( Rationals, CircleObject( 2 ) ); true gap> IsUnit( ZmodnZ(8), CircleObject( ZmodnZObj(2,8) ) ); true gap> m := CircleObject( [ [ 1, 1 ],[ 0, 1 ] ] );; gap> IsUnit( FullMatrixAlgebra( Rationals, 2 ), m ); true

If the first argument is omitted, the result will be returned with respect to the default ring of the circle object `x`.

gap> IsUnit( CircleObject( -2 ) ); true gap> IsUnit( CircleObject( 2 ) ); false gap> IsUnit( CircleObject( ZmodnZObj(2,8) ) ); true gap> IsUnit( CircleObject( [ [ 1, 1 ],[ 0, 1 ] ] ) ); false

`‣ IsCircleUnit` ( [R, ]x ) | ( operation ) |

Let `x` be an element of the ring `R`. Then `IsCircleUnit( R, x )`

determines whether `x` is invertible in `R` with respect to the circle multilpication. This is equivalent to the condition that 1+`x` is a unit in `R` with respect to the ordinary multiplication.

gap> IsCircleUnit( Integers, -2 ); true gap> IsCircleUnit( Integers, 2 ); false gap> IsCircleUnit( Rationals, 2 ); true gap> IsCircleUnit( ZmodnZ(8), ZmodnZObj(2,8) ); true gap> m := [ [ 1, 1 ],[ 0, 1 ] ]; [ [ 1, 1 ], [ 0, 1 ] ] gap> IsCircleUnit( FullMatrixAlgebra(Rationals,2), m ); true

If the first argument is omitted, the result will be returned with respect to the default ring of `x`.

gap> IsCircleUnit( -2 ); true gap> IsCircleUnit( 2 ); false gap> IsCircleUnit( ZmodnZObj(2,8) ); true gap> IsCircleUnit( [ [ 1, 1 ],[ 0, 1 ] ] ); false

`‣ AdjointSemigroup` ( R ) | ( attribute ) |

If `R` is a finite ring then `AdjointSemigroup(`

will return the monoid which is formed by all elements of `R`)`R` with respect to the circle multiplication.

The implementation is rather straightforward and was added to provide a link to the **GAP** functionality for semigroups. It assumes that the enumaration of all elements of the ring `R` is feasible.

gap> R:=Ring( [ ZmodnZObj(2,8) ] ); <ring with 1 generators> gap> S:=AdjointSemigroup(R); <monoid with 4 generators>

`‣ AdjointGroup` ( R ) | ( attribute ) |

If `R` is a finite radical algebra then `AdjointGroup(`

will return the adjoint group of `R`)`R`, given as a group generated by a set of circle objects.

To compute the adjoint group of a finite radical algebra, **Circle** uses the fact that all elements of a radical algebra form a group with respect to the circle multiplication. Thus, the adjoint group of `R` coincides with `R` elementwise, and we can randomly select an appropriate set of generators for the adjoint group.

The warning is displayed by `IsGeneratorsOfMagmaWithInverses`

method defined in `gap4r4/lib/grp.gi`

and may be ignored.

**WARNINGS:**

1. The set of generators of the returned group is not required to be a generating set of minimal possible order.

2. `AdjointGroup`

is stored as an attribute of `R`, so for the same copy of `R` calling it again you will get the same result. But if you will create another copy of `R` in the future, the output may differ because of the random selection of generators. If you want to have the same generating set, next time you should construct a group immediately specifying circle objects that generate it.

3. In most cases, to investigate some properties of the adjoint group, it is necessary first to convert it to an isomorphic permutation group or to a PcGroup.

For example, we can create the following commutative 2-dimensional radical algebra of order 4 over the field of two elements, and show that its adjoint group is a cyclic group of order 4:

gap> x:=[ [ 0, 1, 0 ], > [ 0, 0, 1 ], > [ 0, 0, 0 ] ];; gap> R := Algebra( GF(2), [ One(GF(2))*x ] ); <algebra over GF(2), with 1 generators> gap> RadicalOfAlgebra( R ) = R; true gap> Dimension(R); 2 gap> G := AdjointGroup( R );; gap> Size( R ) = Size( G ); true gap> StructureDescription( G ); "C4"

In the following example we construct a non-commutative 3-dimensional radical algebra of order 8 over the field of two elements, and demonstrate that its adjoint group is the dihedral group of order 8:

gap> x:=[ [ 0, 1, 0 ], > [ 0, 0, 0 ], > [ 0, 0, 0 ] ];; gap> y:=[ [ 0, 0, 0 ], > [ 0, 0, 1 ], > [ 0, 0, 0 ] ];; gap> R := Algebra( GF(2), One(GF(2))*[x,y] ); <algebra over GF(2), with 2 generators> gap> RadicalOfAlgebra(R) = R; true gap> Dimension(R); 3 gap> G := AdjointGroup( R ); <group of size 8 with 2 generators> gap> StructureDescription( G ); "D8"

If the ring `R` is not a radical algebra, then **Circle** will use another approach. We will enumerate all elements of the ring `R` and select those that are units with respect to the circle multiplication. Then we will use a random approach similar to the case of the radical algebra, to find some generating set of the adjoint group. Again, all warnings 1-3 above refer also to this case.

Of course, enumeration of all elements of `R` should be feasible for this computation. In the following example we demonstrate how it works for rings, generated by residue classes:

gap> R := Ring( [ ZmodnZObj(2,8) ] ); <ring with 1 generators> gap> G := AdjointGroup( R ); <group of size 4 with 2 generators> gap> StructureDescription( G ); "C2 x C2" gap> R := Ring( [ ZmodnZObj(2,256) ] ); <ring with 1 generators> gap> G := AdjointGroup( R );; gap> StructureDescription( G ); "C64 x C2"

Due to the `AdjointSemigroup`

(3.3-1), there is also another way to compute the adjoint group of a ring \(R\) by means of the computation of its adjoint semigroup \(S(R)\) and taking the Green's \(H\)-class of the multiplicative neutral element of \(S(R)\). Let us repeat the last example in this way:

gap> R := Ring( [ ZmodnZObj(2,256) ] ); <ring with 1 generators> gap> S := AdjointSemigroup( R ); <monoid with 128 generators> gap> H := GreensHClassOfElement(S,One(S)); {CircleObject( ZmodnZObj( 0, 256 ) )} gap> G:=AsGroup(H); <group of size 128 with 2 generators> gap> StructureDescription(G); "C64 x C2"

However, the conversion of the Green's \(H\)-class to the group may take some time which may vary dependently on the particular ring in question, and will also display a lot of warnings about the default `IsGeneratorsOfMagmaWithInverses`

method, so we did not implemented this as as standard method. In the following example the method based on Green's \(H\)-class is much slower than an application of earlier described random approach (20s vs 10ms):

gap> R := Ring( [ ZmodnZObj(2,256) ] ); <ring with 1 generators> gap> AdjointGroup(R);; gap> R := Ring( [ ZmodnZObj(2,256) ] ); <ring with 1 generators> gap> S:=AdjointSemigroup(R); <monoid with 128 generators> gap> AsGroup(GreensHClassOfElement(S,One(S))); <group of size 128 with 2 generators>

Finally, note that if `R` has a unity \(1\), then the set \(1+R^{ad}\), where \(R^{ad}\) is the adjoint semigroup of `R`, coincides with the multiplicative semigroup \(R^{mult}\) of \(R\), and the map \( r \mapsto (1+r) \) for \(r\) in \(R\) is an isomorphism from \(R^{ad}\) onto \(R^{mult}\).

Similarly, the set \(1+R^*\), where \(R^{*}\) is the adjoint group of `R`, coincides with the unit group of \(R\), which we denote \(U(R)\), and the map \(r \mapsto (1+r)\) for \(r\) in \(R\) is an isomorphism from \(R^*\) onto \(U(R)\).

We demonstrate this isomorphism using the following example.

gap> LoadPackage( "laguna", false ); true gap> FG := GroupRing( GF(2), DihedralGroup(8) ); <algebra-with-one over GF(2), with 3 generators> gap> R := AugmentationIdeal( FG );; gap> G := AdjointGroup( R );; gap> IdGroup( G ); [ 128, 170 ] gap> IdGroup( Units( FG ) ); #I LAGUNA package: Computing the unit group ... [ 128, 170 ]

Thus, dependently on the ring `R`

in question, it might be possible that you can compute much faster its unit group using `Units(R)`

than its adjoint group using `AdjointGroup(R)`

. This is why in an attempt of computation of the adjoint group of the ring with one a warning message will be displayed:

gap> Size( AdjointGroup( GroupRing( GF(2), DihedralGroup(8) ) ) ); WARNING: usage of AdjointGroup for associative ring <R> with one!!! In this case the adjoint group is isomorphic to the unit group Units(<R>), which possibly may be computed faster!!! 128 gap> Size( AdjointGroup( Integers mod 11 ) ); WARNING: usage of AdjointGroup for associative ring <R> with one!!! In this case the adjoint group is isomorphic to the unit group Units(<R>), which possibly may be computed faster!!! 10

If `R` is infinite, an error message will appear, telling that **Circle** does not provide methods to deal with infinite rings.

`‣ InfoCircle` | ( info class ) |

`InfoCircle`

is a special Info class for **Circle** algorithms. It has 2 levels: 0 (default) and 1. To change info level to `k`

, use command `SetInfoLevel(InfoCircle, k)`

.

gap> SetInfoLevel( InfoCircle, 1 ); gap> SetInfoLevel(InfoCircle,1); gap> R := Ring( [ ZmodnZObj(2,8) ]); <ring with 1 generators> gap> G := AdjointGroup( R ); #I Circle : <R> is not a radical algebra, computing circle units ... #I Circle : searching generators for adjoint group ... <group of size 4 with 2 generators> gap> SetInfoLevel( InfoCircle, 0 );

generated by GAPDoc2HTML