88 Coxeter cosets

Let R be a root system in the real vector space V as in Chapter Root systems and finite Coxeter groups. We say that F0GL(V) is an automorphism of R if it is of finite order, it permutes R and the dual F0*GL(V) permutes the coroots R⊂ V. It follows that F0 normalizes the reflection group W associated to R, that is w→ F0wF0-1 is an automorphism F: W→ W. Thus (see Reflection cosets) we get a reflection coset WF0, called here a Coxeter coset.

The motivation for introducing Coxeter cosets comes from non-split reductive algebraic groups over finite fields. Let us, as in the chapter Algebraic groups and semi-simple elements fix a connected reductive algebraic group G over this time an algebraic closure Fq of a finite field Fq; we assume G is defined over Fq, which corresponds to be given a Frobenius endomorphism F so that the finite group of rational points G(Fq) identifies to the subgroup GF of fixed points under F.

Let T be a maximal torus of G, and Φ (resp. Φ) be the roots (resp. coroots) of G with respect to T in the character group X(T) (resp. the group of one-parameter subgroups Y(T)). As explained in Algebraic groups and semi-simple elements then G is determined up to isomorphism by (X(T),Φ,Y(T),Φ) and in CHEVIE this corresponds to give a rational reflection group W=NG(T)/T acting on the vector space V=Q⊗ X(T) together with a root system.

The Frobenius endomorphism F acts also naturally on X(T) and defines thus an endomorphism of V, which is of the form q F0, where F0GL(V) is of finite order and normalizes W. We get thus a Coxeter coset WF0GL(V). The data (X(T),Φ,Y(T),Φ,F0), and the integer q completely determine up to isomorphism the associated reductive finite group GF. Thus we can think of this datum as a way of representing in CHEVIE the essential information which determines a finite reductive group. Indeed, all properties of Chevalley groups can be computed from that datum: symbols representing unipotent characters, conjugacy classes, and finally the whole character table of GF.

It turns out that an interesting part of the objects attached to this datum depends only on (V,W, F0): the order of the maximal tori, the ``fake degrees", the order of GF, Deligne-Lusztig induction in terms of ``almost characters", symbols representing unipotent characters and Deligne-Lusztig induction, etc... (see, e.g., BMM93). It is thus possible to extend their construction to non-crystallographic groups (or even to more general complex reflection groups, see Spetses); this is why we did not include a root system in the definition of a reflection coset. However, unipotent conjugacy classes for instance depend on the root system thus do not exist in general.

We assume now that T is contained in an F-stable Borel subgroup of G. This defines an order on the roots, and there is a unique element φ∈ W F0, reduced element of the coset, which preserves the set of positive roots. This element is stored in the component phi of the coset, and can be used to test isomorphism of cosets. The coset W φ is completely defined by the permutation phi of the roots induced by φ when G is semi-simple (which corresponds to W essential, or to Φ generating V). So in this case we just need to specify phi when defining the coset.

For arbitrary Coxeter cosets, we define phi as follows. Let S be the standard generators so (W,S) is a Coxeter system. Then F0(S) is another set of generators, thus conjugate to S by a unique element of W. It follows that there is a unique element φ∈ WF0 which stabilizes S. We consider from now on cosets of the form where φ stabilizes S.

We should mention a natural generalization which does not exactly fit the above description, and represents in particular the case of the Ree and Suzuki groups. This is when the group is defined as GF where F is not a Frobenius endomorphism, but an isogeny such that some power Fn is a Frobenius endomorphism. Here, F still defines an endomorphism of V which normalizes W; we define a real number q such that Fn is a Frobenius endomorphism attached to an Fqn-structure. Then we still have F=q F0 where F0 is of finite order but q is no more an integer. For instance, for the Ree and Suzuki groups, F0 is an automorphism of order 2 of W, which is of type G2, B2 or F4, and q=√ 2 for B2 and F4 and q=√ 3 for G2 To get this, we need to start from root systems for G2, B2 or F4 where all the roots have the same length. This kind of root system is not crystallographic. Such more general root systems also exist for all finite Coxeter groups such as the dihedral groups and H3 and H4. We will call here Weyl cosets the cosets corresponding to rational forms of algebraic groups, which include thus some non-rational roots systems for B2, G2 and F4.

Conjugacy classes and irreducible characters of Coxeter cosets are defined as for general reflection cosets. For irreducible characters of Weyl cosets, in CHEVIE we choose (following Lusztig) for each φ-stable character of W a particular extension to a character of W⋊⟨ φ ⟩, which we will call the preferred extension. The character table of the coset is the table of the restrictions to of the preferred extensions (See also the section below CharTable for Coxeter cosets). The question of finding the conjugacy classes and character table of a Coxeter coset can be reduced to the case of irreducible root systems R.

• The automorphism φ permutes the irreducible components of W, and is a direct product of cosets where φ permutes cyclically the irreducible components of W. The preferred extension is defined to be the direct product of the preferred extension in each of these situations.

• Assume now that is a descent of scalars, that is the decomposition in irreducible components W=W1× ... × Wk is cyclically permuted by φ. Then there are natural bijections from the φ-conjugacy classes of W to the φk-conjugacy classes of W1 as well as from the φ-stable characters of W to the φk-stable characters of W1, which reduce the definition of preferred extensions on to the definition for W1φk.

• Assume now that W is the Coxeter group of an irreducible root system. φ permutes the simple roots, hence induces a graph automorphism on the corresponding Dynkin diagram. If φ=1 then conjugacy classes and characters coincide with those of the Coxeter group W.

• The nontrivial cases for crystallographic roots systems are (the order of φ is written as left exponent to the type): 2An, 2Dn, 3D4, 2E6.

• For non-crystallographic root systems where all the roots have the same length the additional cases 2B2, 2G2, 2F4 and 2I2(k) arise.

• In case 3D4 the group W⋊⟨ φ⟩ can be embedded into the Coxeter group of type F4, which induces a labeling for the conjugacy classes of the coset. The preferred extension is chosen as the (single) extension with rational values.

• In case 2Dn the group W⋊⟨ φ⟩ is isomorphic to a Coxeter group of type Bn. This induces a canonical labeling for the conjugacy classes of the coset and allows to define the preferred extension in a combinatorial way using the labels (pairs of partitions) for the characters of the Coxeter group of type Bn.

• In the remaining crystallographic cases φ identifies to -w0 where w0 is the longest element of W. So, there is a canonical labeling of the conjugacy classes and characters of the coset by those of W. The preferred extensions are defined by describing the signs of the character values on -w0.

In GAP we construct the Coxeter coset by starting from a Coxeter datum specified by the matrices of simpleRoots and simpleCoroots, and giving in addition the matrix F0Mat of the map F0:V→ V (see the commands CoxeterCoset and CoxeterSubCoset). As it is true for the Coxeter groups the elements of are uniquely determined by the permutation they induce on the set of roots R. We consider these permutations as Elements of the Coxeter coset.

Coxeter cosets are implemented in GAP by a record which points to a Coxeter datum record and has additional fields holding F0Mat and the corresponding element phi. Functions on the coset work with elements of the group coset W φ (for example, ChevieClassInfo); however, most definitions for elements of untwisted Coxeter groups apply without change to elements in W φ: e.g., if we define the length of an element wφ∈ W φ as the number of positive roots it sends to negative ones, it is the same as the length of w, i.e., φ is of length 0, since φ has been chosen to preserve the set of positive roots. Similarly, the CoxeterWord describing is the same as the one for w, etc...

We associate to a Coxeter coset a twisted Dynkin diagram, consisting of the Dynkin diagram of W and the graph automorphism induced by φ on this diagram (this specifies the group W⋊⟨ F⟩, mentioned above, up to isomorphism). See the functions ReflectionType, ReflectionName and PrintDiagram for Coxeter cosets.

Below is an example showing first how to not define, then how to define, the Weyl coset for a Suzuki group:

   gap> 2B2:=CoxeterCoset(CoxeterGroup("B",2),(1,2));
   Error, #I transposed of matrix for F0 must normalize set of coroots of parent.
   in
   CoxeterCoset( CoxeterGroup( "B", 2 ), (1,2) ) called from
   main loop
   brk>
   gap> 2B2:=CoxeterCoset(CoxeterGroup("Bsym",2),(1,2));
   2Bsym2
   gap> Display(CharTable(2B2));
   2Bsym2

        2 1      2     2

           	 1   121

   2.     1      1     1
   .11    1     -1    -1
   1.1    . -ER(2) ER(2)

A subcoset Hwφ of is given by a reflection subgroup H of W and an element w of W such that induces an automorphism of the root system of H. For algebraic groups, this corresponds to a rational form of a reductive subgroup of maximal rank. For example, if corresponds to the algebraic group G and H is the trivial subgroup, the coset Hwφ corresponds to a maximal Torus Tw of type w.

   gap> CoxeterSubCoset(2B2,[],Group(2B2).1);
   (q^2-ER(2)q+1)

A subgroup H which is a parabolic subgroup corresponds to a rational form of a Levi subgroup of G. The command Twistings gives all rational forms of such a Levi.

   gap> W:=CoxeterGroup("B",2);
   CoxeterGroup("B",2)
   gap> Twistings(W,[1]);  
   [ ~A1.(q-1), ~A1.(q+1) ]
   gap> Twistings(W,[2]);
   [ A1<2>.(q-1), A1<2>.(q+1) ]

Notice how we distinguish between subgroups generated by short roots and by long roots. A general H corresponds to a reductive subgroup of maximal rank. Here we consider the subgroup generated by the long roots in B2, which corresponds to a subgroup of type SL2× SL2 in SP4, and show its possible rational forms.

   gap> W:=CoxeterGroup("B",2);
   CoxeterGroup("B",2)
   gap> Twistings(W,[2,4]);
   [ A1<2>xA1<4>, (A1xA1)<2,4> ]

Subsections

  1. CoxeterCoset
  2. CoxeterSubCoset
  3. Functions on Coxeter cosets
  4. ReflectionType for Coxeter cosets
  5. ChevieClassInfo for Coxeter cosets
  6. CharTable for Coxeter cosets
  7. Frobenius
  8. Twistings for Coxeter cosets

88.1 CoxeterCoset

CoxeterCoset( W[, FMat ] )

CoxeterCoset( W[, FPerm] )

This function returns a Coxeter coset as a GAP object. The argument W must be a Coxeter group (created by CoxeterGroup or ReflectionSubgroup). In the first form the argument F0Mat must be an invertible matrix of dimension Rank(W), representing an automorphism F of the root system of the parent of W. In the second form FPerm is a permutation of the roots of the parent group of W; it is assumed that the corresponding FMat acts trivially on the orthogonal of these roots. A shortcut is accepted if W has no parent or same rank as its parent and FPerm preserve its simple roots: one need only give the induced permutation of the simple roots.

If there is no second argument the default for F0Mat is the identity matrix.

CoxeterCoset returns a record from which we document the following components:

isDomain, isFinite:

true

reflectionGroup:

the Coxeter group W

F0Mat:

the matrix acting on V which represents the unique element φ in WF0 which preserves the positive roots.

phi:

the permutation on the roots of W induced by F0Mat (the element of smallest length in the coxeter Coset).

In the first example we create a Coxeter coset corresponding to the general unitary groups GU3(q) over finite fields with q elements.

    gap> W := RootDatum("gl",3);;
    gap> gu3 := CoxeterCoset( W, -IdentityMat( 3 ) );
    2A2.(q+1)
    gap> F4 := CoxeterGroup( "F", 4 );;
    gap> D4 := ReflectionSubgroup( F4, [ 1, 2, 16, 48 ] );;
    gap> PrintDiagram( D4 );
    D4   9
          \
           1 - 16
          /
         2
    gap> CoxeterCoset( D4, MatXPerm(D4,(2,9,16)) );
    3D4<9,16,1,2>
    gap> CoxeterCoset( D4, (2,9,16));
    3D4<9,16,1,2>

These functions require the package "chevie" (see RequirePackage).

88.2 CoxeterSubCoset

CoxeterSubCoset( WF, r, [w] )

Returns the reflection subcoset of the Coxeter coset WF generated by the reflections with roots specified by r. r is a list of indices specifying a subset of the roots of W where W is the Coxeter group CoxeterGroup(WF). If specified, w must be an element of W such that w*WF.phi normalizes the subroot system generated by r. If absent, the default value for w is (). It is an error, if w*WF.phi does not normalize the subsystem.

    gap> CoxeterSubCoset( CoxeterCoset( CoxeterGroup( "A", 2 ), (1,2) ),
    >                                                              [ 1 ] );
    Error, must give w, such that w * WF.phi normalizes subroot system.
     in
    CoxeterSubCoset(CoxeterCoset(CoxeterGroup("A",2), (1,2)), [ 1 ])
     called from main loop
    brk>
    gap> f4coset := CoxeterCoset( CoxeterGroup( "F", 4 ) );
    F4
    gap> w := RepresentativeOperation( CoxeterGroup( f4coset ),
    >                      [ 1, 2, 9, 16 ], [ 1, 9, 16, 2], OnTuples );;
    gap> 3d4again := CoxeterSubCoset( f4coset, [ 1, 2, 9, 16], w );
    3D4<9,16,1,2>
    gap> PrintDiagram( 3d4again );
    phi acts as ( 2, 9,16) on the component below
    D4   9
          \
           1 - 2
          /
         16 

This function requires the package "chevie" (see RequirePackage).

88.3 Functions on Coxeter cosets

All functions for reflection cosets are implemented for Coxeter cosets.

This includes Group( WF ) which returns the Coxeter group of which WF is a coset; the functions Elements, Random, Representative, Size, in, which use the corresponding functions for Group( WF ); ConjugacyClasses( WF ), which returns the φ-conjugacy classes of W; the corresponding PositionClass( WF , x ) and FusionConjugacyClasses( HF, WF ), InductionTable( HF, WF ).

For Weyl coset associated to a finite reductive group GF, the characters of the coset correspond to unipotent Deligne-Lusztig characters, and the induction from a subcoset corresponding to a Levi subgroup corresponds to the Lusztig induction of the Deligne-Lusztig characters (see more details in the next chapter). Here are some examples:\

Harish-Chandra induction in the basis of almost characters:

    gap> WF := CoxeterCoset( CoxeterGroup( "A", 4 ), (1,4)(2,3) );
    2A4
    gap> Display( InductionTable( CoxeterSubCoset( WF, [ 2, 3 ] ), WF ) );
    Induction from 2A2<2,3>.(q-1)(q+1) to 2A4
          |111 21 3
    ________________
    11111 |  1  . .
    2111  |  .  1 .
    221   |  1  . .
    311   |  1  . 1
    32    |  .  . 1
    41    |  .  1 .
    5     |  .  . 1

Lusztig induction from a diagonal Levi:

    gap> HF := CoxeterSubCoset( WF, [1, 2],
    >                LongestCoxeterElement( CoxeterGroup( WF ) ) );;
    gap> Display( InductionTable( HF, WF ) );
    Induction from 2A2.(q+1)^2 to 2A4
          |111 21  3
    _________________
    11111 | -1  .  .
    2111  | -2 -1  .
    221   | -1 -2  .
    311   |  1  2 -1
    32    |  . -2  1
    41    |  .  1 -2
    5     |  .  .  1

A descent of scalars:

    gap> W := CoxeterCoset( CoxeterGroup( "A", 2, "A", 2 ), (1,3)(2,4) );
    (A2xA2)
    gap> Display( InductionTable( CoxeterSubCoset( W, [ 1, 3 ] ), W ) );
    Induction from (A1xA1)<1,3>.(q-1)(q+1) to (A2xA2)
        |11 2
    __________
    111 | 1 .
    21  | 1 1
    3   | . 1 

Print( WF ), ReflectionName( WF ) and PrintDiagram( WF ) show the isomorphism type of the reductive group GF. An orbit of φ=WF.phi on the components is put in brackets if of length k greater than 1, and is preceded by the order of φk on it, if this is not 1. For example "2(A2xA2)" denotes 2 components of type A2 permuted by φ, and such that φ2 induces the non-trivial diagram automorphism on any of them, while 3D4 denotes an orbit of length 1 on which φ is of order 3.

    gap> W := CoxeterCoset( CoxeterGroup( "A", 2, "G", 2, "A", 2 ),
    >                                                       (1,5,2,6) );
    2(A2xA2)<1,2,5,6>xG2<3,4>
    gap> ReflectionName( W );
    "2(A2xA2)<1,2,5,6>xG2<3,4>"
    gap> W := CoxeterCoset( CoxeterGroup( "A", 2, "A", 2 ), (1,3,2,4) );
    2(A2xA2)
    gap> PrintDiagram( W );
    phi permutes the next 2 components
    phi^2 acts as (1,2) on the component below
    A2    1 - 2
    A2    3 - 4 

ChevieClassInfo( WF ), see the explicit description in ChevieClassInfo for Coxeter cosets.

ChevieCharInfo, CharParams, CharName return additional information on the irreducible characters.

Finally, some functions for elements of a Coxeter group work naturally for elements of a Coxeter coset: CoxeterWord, EltWord, CoxeterLength, LeftDescentSet, RightDescentSet, ReducedInRightCoset, etc... These functions take the same value on wφ in Wφ that they take on w∈ W.

88.4 ReflectionType for Coxeter cosets

ReflectionType( WF )

returns the type of the Coxeter coset WF. This consists of a list of records, one for each orbit of WF.phi on the irreducible components of the Dynkin diagram of CoxeterGroup(WF), which have two fields:

orbit:

is a list of types of the irreducible components in the orbit. These types are the same as returned by the function ReflectionType for an irreducible untwisted Coxeter group (see ReflectionType in chapter Root systems and finite Coxeter groups): a couple [type,indices] (a triple for type I2(n)). The components are ordered according to the action of WF.phi, so WF.phi maps the generating permutations with indices in the first type to indices in the second type in the same order as stored in the type, etc ...
phi:
if k is the number of irreducible components in the orbit, this is the permutation which describes the action of WF.phik on the simple roots of the first irreducible component in the orbit.

    gap> W := CoxeterCoset( CoxeterGroup( "A", 2, "A", 2 ), (1,3,2,4) );
    2(A2xA2)
    gap> ReflectionType( W );
    [ rec(
          orbit := [rec(
                  rank    :=2,
                  series  :="A",
                  indices :=[ 1, 2 ]), rec(
                  rank    :=2,
                  series  :="A",
                  indices :=[ 3, 4 ]) ],
          twist :=(1,2)) ]

This function requires the package "chevie" (see RequirePackage).

88.5 ChevieClassInfo for Coxeter cosets

ChevieClassInfo( WF )

returns information about the conjugacy classes of the Coxeter coset WF. The result is a record with three components: classtext contains a list of reduced words for the representatives in ConjugacyClasses(WF), classnames contains corresponding names for the classes, and classparams gives corresponding parameters for the classes. Let W be the Coxeter group CoxeterGroup(WF). In the case where -1∉ W, i.e., φ=-w0, they are obtained by multiplying by w0 a set of representatives of maximal length of the classes of W.

    gap> W := CoxeterGroup( "D", 4 );;
    gap> ChevieClassInfo( CoxeterCoset( W, (1,2,4) ) );
    rec(
      classtext := [ [ 1 ], [  ], [ 1, 2, 3, 1, 2, 3 ], [ 3 ], [ 1, 3 ],
          [ 1, 2, 3, 1, 2, 4, 3, 2 ], [ 1, 2, 3, 2 ] ],
      classnames :=
       [ "C_3", "\\tilde A_2", "C_3+A_1", "\\tilde A_2+A_1", "F_4", 
         "\\tilde A_2+A_2", "F_4(a_1)" ],
      classparams :=
       [ [ "C_3" ], [ "~A_2" ], [ "C_3+A_1" ], [ "~A_2+A_1" ],
         [ "F_4" ], [ "~A_2+A_2" ], [ "F_4(a_1)" ] ] )

This function requires the package "chevie" (see RequirePackage).

88.6 CharTable for Coxeter cosets

CharTable( WF )

This function returns the character table of the Coxeter coset WF (see also the introduction of this Chapter). We call ``characters'' of the Coxeter coset WF with corresponding Coxeter group W the restriction to W φ of a set containing one extension of each φ-invariant character of W to the semidirect product of W with the cyclic group generated by φ. (We choose, following Lusztig, in each case one non-canonical extension, called the preferred extension.)

The returned record contains almost all components present in the character table of a Coxeter group. But if φ is not trivial then there are no components powermap (since powers of elements in the coset need not be in the coset) and orders (if you really need them, use MatXPerm to determine the order of elements in the coset).

    gap> W := CoxeterCoset( CoxeterGroup( "D", 4 ), (1,2,4) );
    CoxeterCoset(CoxeterGroup("D",4), (1,2,4))
    gap> Display( CharTable( W ) );
    3D4

           2  2   2     2      2  2      3      3
           3  1   1     1      .  .      1      1

             C3 ~A2 C3+A1 ~A2+A1 F4 ~A2+A2 F4(a1)

    .4        1   1     1      1  1      1      1
    .1111    -1   1     1     -1  1      1      1
    .22       .   2     2      . -1     -1     -1
    11.2      .   .     .      . -1      3      3
    1.3       1   1    -1     -1  .     -2      2
    1.111    -1   1    -1      1  .     -2      2
    1.21      .   2    -2      .  .      2     -2

This function requires the package "chevie" (see RequirePackage).

88.7 Frobenius

Frobenius( WF )( o [, i])

Given a Coxeter coset WF, Frobenius(WF) returns a function which makes WF.phi act on its argument which is some object for which this action has been defined. If o is a list, it applies recursively Frobenius to each element of the list. If it is a permutation or an integer, it returns o^(WF.phi^-1). If a second argument i is given, it applies Frobenius raised to the i-th power (this is convenient for instance to apply the inverse of Frobenius). Finally, for an arbitrary object defined by a record, it looks if the method o.operations.Frobenius is defined and if so calls this function with arguments WF, o and i (with i=1 if it was omitted).

Such an action of the Frobenius is defined for instance for braids, Hecke elements and semisimple elements.

    gap> W:=CoxeterGroup("E",6);;WF:=CoxeterCoset(W,(1,6)(3,5));
    2E6
    gap> T:=Basis(Hecke(Group(WF)),"T");;Frobenius(WF)(T(1));
    T(6)
    gap> B:=Braid(W);; Frobenius(WF)(B(1,2,3));
    265
    gap> s:=SemisimpleElement(W,[1..6]/6);
    <1/6,1/3,1/2,2/3,5/6,0>
    gap> Frobenius(WF)(s);
    <0,1/3,5/6,2/3,1/2,1/6>
    gap> W:=CoxeterGroup("D",4);WF:=CoxeterCoset(W,(1,2,4));
    CoxeterGroup("D",4)
    3D4
    gap> B:=Braid(W);;b:=B(1,3);
    13
    gap> Frobenius(WF)(b);
    43
    gap> Frobenius(WF)(b,-1);
    23

This function requires the package "chevie" (see RequirePackage).

88.8 Twistings for Coxeter cosets

Twistings( W, L )

W should be a Coxeter group record or a Coxeter coset record, and L should be a reflection subgroup of W (or of Group(W) for a coset), or a sublist of the generating reflections of W (resp. Group(W)), in which case the call is the same as Twistings(W,ReflectionSubgroup(W,L)) (resp. Twistings(W,ReflectionSubgroup(Group(W),L))).

The function returns the list, up to W-conjugacy, of Coxeter sub-cosets of W whose Coxeter group is L --- In term of algebraic groups, it corresponds to representatives of the possible twisted forms of the reductive subgroup of maximal rank L. In the case that W represents a coset , the subgroup L must be conjugate to φ(L) for a rational form to exist. If normalizes L, then the rational forms are classified by the the φ-classes of NW(L)/L.

    gap> W:=CoxeterGroup("E",6);
    CoxeterGroup("E",6)
    gap> WF:=CoxeterCoset(W,(1,6)(3,5));
    2E6
    gap> L:=ReflectionSubgroup(W,[2..5]);
    ReflectionSubgroup(CoxeterGroup("E",6), [ 2, 3, 4, 5 ])
    gap> Twistings(W,L);
    [ D4<2,3,4,5>.(q-1)^2, 3D4<2,3,4,5>.(q^2+q+1), 2D4<2,3,4,5>.(q-1)(q+1) ]
    gap> Twistings(WF,L);
    [ 2D4<2,5,4,3>.(q-1)(q+1), 3'D4<2,5,4,3>.(q^2-q+1), D4<2,3,4,5>.(q+1)^2]

This function requires the package "chevie" (see RequirePackage). Previous Up Next
Index

GAP 3.4.4
April 1997