<!-- %W  maintain.xml    GAP 4 package AtlasRep             Thomas Breuer -->


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Chapter Label="chap:maintain">
<Heading>Maintenance Issues of the &AtlasRep; Package</Heading>

This is just preliminary, in particular not intended for inclusion in the manual!

<hier/>

This chapter describes why some data that are available in the web &ATLAS;
are excluded from the &GAP; interface,
by which data they have been replaced (if applicable),
and which additional data are distributed together with the
&AtlasRep; package.


-> section on just added material, such as cyc2ccls scripts

-> document all in datapkg!


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Section Label="sect:kernel_scripts">
<Heading>Generators of Kernels</Heading>


- change the code!
  - test.g -> not only for compatible std. generators & same std.!

run for 3^(1+12):2.Suz.2
Print( "hier\n" );
if entry[1] = "3^(1+12):2.Suz.2" then
  SetSize( N, 3^13 );
fi;
#T  for 3^13, calling Size exceeds the memory!
#T  (trans. permutation action on 3^12 = 531441 points!)

-> how to verify that the kernel is correct?


Note that the program computes elements in G,
in particular we are not concerned with standardization in F,
and we do not assume that the given standard generators of G
are compatible with some standard generators of F.
(Only relation: If G has several normal subgroups N such that G/N
is isomorphic with F and if the i-th std. gens of G and F are compatible
then we prefer the normal subgroup that is the kernel of the epim.
given by mapping the i-th std. gen. of G to those of F.)


- provide kernel info also for incompatible generators:

AGR.STDCOMP("(A5xA12):2",[0,"A12.2",1,false]);
AGR.STDCOMP("(A5xU3(8):3):2",[0,"A5.2",1,false]);
AGR.STDCOMP("(L3(2)xS4(4):2).2",[0,"L3(2).2",1,false]);
AGR.STDCOMP("2^2.2E6(2).S3",[0,"2E6(2).3.2",0,false]);
AGR.STDCOMP("5^3.L3(5)",[2,"L3(5)",1,false]);

- a new type of slps, many new scripts

- hard case:
  2.2E6(2) ->> 2E6(2) (and analog. 2^2.2E6(2) -> 2E6(2))
  need to skip obviously unnecessary words;
  the > 5593818-th word works (which exactly?),
  but only 525 words must actually be tested (really?)
  (altogether 10 minutes runtime!)

  (done 2015-04-13 on gemma)

5593773-
5593781+
5593782+
5593813-
5593814+
5593818-
[ [ m2^2*m1*m2*(m2*m1*m2^2*m1)^2*(m2*m1)^4, 33 ], true ]

-> is equal to (m2^2*m1)^3*m2*m1*m2^2*m1*(m2*m1)^4
   with slp 3:= 2*1
            4:= 2*3
            5:= 4^3  (2 mult)
            6:= 3^4  (2 mult)
            7:= 5*3
            8:= 7*4
            9:= 8*6
   thus 9 mult!
-> add a remark about the arbitr. of the printed factoriz.!

(only those numbers are shown for which all syllables are below the el. order,
and "-" means that the order in the factor group is even, which need not be
checked)

- function is AGR.Test.ComputeKernelGenerators...

-> better move out from Test!

- really verify the non-cyclic kernels! (see TODO_2)



--------------------------------------------------------------------------

kernel generators:

- missing verifications in atlasrep/kerrun.out? (std. 0)
- two kernels of order 3^13 --prove!

# run on 2014-04-30

# add a test:
# if a repres. for G and of m.G are avail. then try to compute kernel slp,
# also if no compat. is assumed;
# try also compatibility ...

# add a test: are all files in the datapkg dir. of atlasrep valid?

......................................................................


# find a kernel generator:

# see AGR.Test.ComputeKernelGenerators! (atlasrep/gap/test.g)

fgens:= AtlasGenerators( "L3(7).2", 1 ).generators;
gens:= AtlasGenerators( "3.L3(7).2", 1 ).generators;


kergens:= [];
kerwords:= [];

f:= FreeMonoid( 2 );
mgens:= GeneratorsOfMonoid( f );
iter:= Iterator( f );
for word in iter do
  m:= MappedWord( word, mgens, gens );
  fm:= MappedWord( word, mgens, fgens );
  ord:= Order( fm );
  if Order( m ) <> ord then
kergen:= m^ord;
if not kergen in kergens then
  Add( kergens, kergen );
  Add( kerwords, word );
  if Length( kerwords ) >= 1 then
Error("!");
  fi;
fi;
  fi;
od;

word;
Order( m );
Order( fm );

brk> word;
m1
brk> Order( m );
4
brk> Order( fm );
2


m2^2*m1*m2*m1



slp:= StraightLineProgram(
[[2,1,1,1],[3,2],[2,2],[5,1,4,1],[[6,19]]],2
);

-----------------------------------------------------------------------------

#I  AGR.Test.KernelGenerators for 2.A10.2: missing kernels of epim. to
#I  [ "A10.2" ]
#I  AGR.Test.KernelGenerators for 2.A12.2: missing kernels of epim. to
#I  [ "A12.2" ]
#I  AGR.Test.KernelGenerators for 2.A13.2: missing kernels of epim. to
#I  [ "A13.2" ]
#I  AGR.Test.KernelGenerators for 2.A8.2: missing kernels of epim. to
#I  [ "A8.2" ]
#I  AGR.Test.KernelGenerators for 2.A9.2: missing kernels of epim. to
#I  [ "A9.2" ]
#I  AGR.Test.KernelGenerators for 2.B: missing kernels of epim. to
#I  [ "B" ]

#I  AGR.Test.KernelGenerators for 2.L2(25): missing kernels of epim. to
#I  [ "L2(25)" ]
#I  AGR.Test.KernelGenerators for 2.O7(3).2: missing kernels of epim. to
#I  [ "O7(3).2" ]
#I  AGR.Test.KernelGenerators for 2.O8+(2): missing kernels of epim. to
#I  [ "O8+(2)" ]
#I  AGR.Test.KernelGenerators for 2.O8+(2).2: missing kernels of epim. to
#I  [ "O8+(2).2" ]

#I  AGR.Test.KernelGenerators for 2.U6(2).2: missing kernels of epim. to
#I  [ "U6(2).2" ]
omit 2^2.2E6(2)
#I  AGR.Test.KernelGenerators for 2^2.2E6(2).S3: missing kernels of epim. to
#I  [ "2E6(2).3.2" ]
#I  AGR.Test.KernelGenerators for 2^2.L3(4): missing kernels of epim. to
#I  [ "2.L3(4)", "L3(4)" ]
#I  AGR.Test.KernelGenerators for 2^2.L3(4).2_2: missing kernels of epim. to
#I  [ "L3(4).2_2" ]

#I  AGR.Test.KernelGenerators for 6.O7(3).2: missing kernels of epim. to
#I  [ "2.O7(3).2", "3.O7(3).2", "O7(3).2" ]

#I  AGR.Test.KernelGenerators for Isoclinic(12.M22.2): missing kernels of epim\
. to
#I  [ "2.M22.2", "3.M22.2", "6.M22.2", "M22.2" ]
#I  AGR.Test.KernelGenerators for Isoclinic(2.A8.2): missing kernels of epim. \
to
#I  [ "A8.2" ]
#I  AGR.Test.KernelGenerators for Isoclinic(2.HS.2): missing kernels of epim. \
to
#I  [ "HS.2" ]
#I  AGR.Test.KernelGenerators for Isoclinic(2.Suz.2): missing kernels of epim.\
 to
#I  [ "Suz.2" ]
#I  AGR.Test.KernelGenerators for Isoclinic(4.M22.2): missing kernels of epim.\
 to
#I  [ "2.M22.2", "M22.2" ]

#I  AGR.Test.KernelGenerators for Isoclinic(6.M22.2): missing kernels of epim.\
 to
#I  [ "3.M22.2", "M22.2" ]

-> no repres.

#I  AGR.Test.KernelGenerators for Isoclinic(6.Suz.2): missing kernels of epim.\
 to
#I  [ "3.Suz.2", "Suz.2" ]

................
#I  AGR.Test.KernelGenerators for 3^(1+12):2.Suz.2: missing kernels of epim. t\
o
#I  [ "2.Suz.2" ]
#I  AGR.Test.KernelGenerators for 3^(1+12):6.Suz.2: missing kernels of epim. t\
o
#I  [ "3^(1+12).2.Suz.2", "6.Suz.2" ]

-> Vorsicht:
   Ich habe zwar gezeigt, dass die Konjugierten des 3. Erzeugers
   mindestens eine 3^13 erzeugen, aber ist es nicht vielleicht mehr?
   Und ich habe nicht gezeigt, dass die Untergruppe wirklich normal ist!

(in beiden Fällen, M3max7G0-ker6Suzd2W1 und Mmax7G0-ker2Suzd2W1)


gap> List( gens.generators, Order );
[ 4, 3, 3 ]
gap> List( fgens.generators, Order );
[ 4, 3 ]
gap> p:= Product( gens.generators{[1,2]} );;
gap> l:= List( [0..11], x -> gens.generators[3]^(p^x) );;
gap> cc:= Group(l );;
gap> orb:= Orbit( cc, l[1][38] );;  Length( orb );
1
gap> orb:= Orbit( cc, l[1][78] );;  Length( orb );
1594323
gap> 3^13;
1594323



g:= Group( gens.generators );
kergens:= [];
kerwords:= [];

f:= FreeMonoid( 2 );
mgens:= GeneratorsOfMonoid( f );
iter:= Iterator( f );
for word in iter do
  m:= MappedWord( word, mgens, gens.generators );
  ord:= Order( m );
  if ord mod 7 = 0 then
kergen:= m^(ord/7);
    if not kergen in kergens then
      n:= NormalClosure( g, SubgroupNC( g, [ kergen ] ) );
      if 7^5 mod Size( n ) = 0 then
        Add( kergens, kergen );
        Add( kerwords, [ word, ord/7 ] );
        if Length( kerwords ) >= 1 then
Error("!");
        fi;
      fi;
    fi;
  fi;
od;

</Section>

ab hier o.k.!


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Section Label="sect:excluded_data">
<Heading>Excluded data files</Heading>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="subsect:3L37d2_wrong_repres">
<Heading>A matrix representation of <M>3.L_3(7).2</M></Heading>

The files <C>3L37d2G1-f7r6B0.m1</C> and <C>3L37d2G1-f7r6B0.m2</C>
that are available in the web &ATLAS;
contain generators for the group <M>G = 3.L_3(7).2</M>,
but these generators are not standard.
First we show this fact and then we compute standard generators.

<P/>

The files look as follows.

<P/>

<Listing>
 1     7     6     6
000454
000103
000163
321000
016000
622000
</Listing>

and

<Listing>
 1     7     6     6
000500
000416
000064
251000
212000
203000
</Listing>

Standard generators of <M>G</M> are defined as follows.

<P/>

<Q>Std. gens. of <M>L_3(7).2</M> are <M>c, d</M>, where <M>c</M> in <C>2B</C>,
<M>d</M> in <C>4B</C> (two outer classes),
<M>|cd| = 19</M>, <M>|cdcdd| = 8</M>.
Std. gens. of <M>3.L_3(7).2</M> are preimages <M>C, D</M>
where <M>|CD| = 19</M>.</Q>

<P/>

We create the matrices in &GAP; and check the conditions.

<P/>

<Example>
gap> gens:= List( [
> " 1     7     6     6\n\
> 000454\n\
> 000103\n\
> 000163\n\
> 321000\n\
> 016000\n\
> 622000",
> " 1     7     6     6\n\
> 000500\n\
> 000416\n\
> 000064\n\
> 251000\n\
> 212000\n\
> 203000" ], str -> ScanMeatAxeFile( str, 7, "string" ) );
[ < immutable compressed matrix 6x6 over GF(7) >, 
  < immutable compressed matrix 6x6 over GF(7) > ]
gap> List( gens, Order );
[ 2, 4 ]
gap> Order( gens[1] * gens[2] );
19
gap> Order( ( gens[1] * gens[2] )^2 * gens[2] );
6
</Example>

<P/>

This shows that the given matrices are not <E>standard</E> generators
of <M>G</M>.
In order to convince ourselves that they generate <M>G</M>,
we proceed as follows.
First we compute a faithful permutation representation of the group <M>H</M>,
say, that is generated by the given matrices.
Then we show that the derived subgroup <M>D</M> of <M>H</M> is a perfect group
with a central subgroup <M>Z</M> of order three such that the factor group
<M>D / Z</M> is a simple group that is isomorphic with <M>L_3(7)</M>,
thus <M>D</M> is the triple cover of <M>L_3(7)</M>.
It remains to show that <M>H / Z</M> is not a direct product of <M>D / Z</M>
and a group of order two; for that, it is enough to show that the centralizer
of an element of order <M>19</M> in <M>H</M> has odd order.

<!--
Note that if <M>x</M> is an element of order <M>19</M> in <M>H</M> and
<M>H / Z</M> is a direct product of <M>D / Z</M> and a group of order two
that is generated by <M>y Z</M> then <M>x^y</M> lies in <M>x Z</M>;
since <M>x</M> is the only element in <M>x Z</M> that has order <M>19</M>,
we have found an element of even order that commutes with <M>x</M>.
-->

<P/>

<Example>
gap> g:= GroupWithGenerators( gens );;
gap> Size( g );
11261376
gap> orbs:= Orbits( g, Elements( GF(7)^6 ) );;
gap> Collected( List( orbs, Length ) );
[ [ 1, 1 ], [ 684, 1 ], [ 16416, 1 ], [ 16758, 6 ] ]
gap> orb:= First( orbs, x -> Length( x ) = 684 );;
gap> acthom:= ActionHomomorphism( g, orb, OnRight );;
gap> img:= Image( acthom );;
gap> Size( img ) = Size( g );
true
gap> der:= DerivedSubgroup( img );;
gap> IsPerfectGroup( der );
true
gap> z:= Centre( der );;
gap> Size( z );
3
gap> f:= der / z;;
gap> IsSimple( f );
true
gap> IsomorphismTypeInfoFiniteSimpleGroup( f );
rec( name := "A(2,7) = L(3,7) ", parameter := [ 3, 7 ], series := "L" )
gap> gensimgs:= List( gens, x -> x^acthom );;
gap> x:= gensimgs[1] * gensimgs[2];;
gap> Order( x );
19
gap> Size( Centralizer( img, x ) );
57
</Example>

<P/>

Now let us find standard generators for <M>G</M>.
According to <Cite Key="CCN85" Where="p. 51"/>,
there is exactly one conjugacy class of elements of the orders <M>2</M>
and <M>4</M> in <M>G</M> outside <M>D</M>,
and the given generators have the right orders and lie outside <M>D</M>.
Thus we may keep the first matrix and replace the second one
by a suitable <M>G</M>-conjugate.

<P/>

<Example>
gap> c:= gensimgs[1];;
gap> d:= gensimgs[2];;
gap> repeat
> dr:= d^Random( img );
> until Order( c * dr ) = 19 and Order( (c * dr)^2 * dr ) = 8
>       and Size( img ) = Size( SubgroupNC( img, [ c, dr ] ) );
gap> 
gap> stdgens:= [ gens[1], PreImagesRepresentative( acthom, dr ) ];
[ < immutable compressed matrix 6x6 over GF(7) >, 
  < immutable compressed matrix 6x6 over GF(7) > ]
</Example>

<P/>

The erroneous representation gets excluded from the &GAP; interface
by removing it from the data list in the file <F>gap/atlasprm.json</F>
of the &AtlasRep; package and then adding an entry to the global variable
<Ref Var="AGR.ExclusionList"/>.

<!-- and add entry to htm/data/changes.htm -->

<P/>

The standard generators computed as shown above have been added to the
data that are distributed together with the &AtlasRep; package,
the representation is now available with the name <C>3L37d2G1-f7r6aB0</C>;
note that the name <C>3L37d2G1-f7r6B0</C> must be avoided.

<P/>

<Example>
gap> OneAtlasGeneratingSetInfo( "3.L3(7).2", Dimension, 6, Ring, GF(7) );
rec( dim := 6, groupname := "3.L3(7).2", id := "a", 
  identifier := 
    [ "3.L3(7).2", 
      [ [ "internal", "3L37d2G1-f7r6aB0.m1" ], 
          [ "internal", "3L37d2G1-f7r6aB0.m2" ] ], 1, 7 ], 
  repname := "3L37d2G1-f7r6aB0", repnr := 1, ring := GF(7), size := 11261376, 
  standardization := 1, type := "matff" )
</Example>

</Subsection>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="subsect:2O73d2_wrong_repres">
<Heading>A matrix representation of <M>2.O_7(3).2^*</M></Heading>

The files <C>2O73d2G1-f3r8B0.m1</C> and <C>2O73d2G1-f3r8B0.m2</C>
that are available in the web &ATLAS;
do not contain generators for the group <M>2.O_7(3).2</M>.

<P/>

The files look as follows.

<P/>

<Listing>
 1     3     8     8
01200212
10011201
20201110
02122011
01111002
00211210
20212110
00111021
</Listing>

and

<Listing>
 1     3     8     8
11010011
10121000
20000221
12101200
11001011
22000112
12121220
01201002
</Listing>

Standard generators of <M>2.O_7(3).2</M> are defined as follows.

<P/>

<Q>Standard generators of <M>O_7(3).2</M> are <M>c, d</M>
where <M>c</M> is in class <C>2D</C>,
<M>d</M> has order <M>7</M>,
<M>cd</M> has order <M>26</M> and <M>cdcdd</M> has order <M>14</M>.
Standard generators of <M>2.O_7(3).2</M> are preimages <M>C, D</M>
where <M>D</M> has order <M>7</M>.</Q>

<P/>

We create the matrices in &GAP; and check the conditions.

<P/>

<Example>
gap> gens:= List( [
> " 1     3     8     8\n\
> 01200212\n\
> 10011201\n\
> 20201110\n\
> 02122011\n\
> 01111002\n\
> 00211210\n\
> 20212110\n\
> 00111021",
> " 1     3     8     8\n\
> 11010011\n\
> 10121000\n\
> 20000221\n\
> 12101200\n\
> 11001011\n\
> 22000112\n\
> 12121220\n\
> 01201002" ], str -> ScanMeatAxeFile( str, 3, "string" ) );
[ < immutable compressed matrix 8x8 over GF(3) >,
  < immutable compressed matrix 8x8 over GF(3) > ]
gap> List( gens, Order );
[ 2, 7 ]
gap> Order( gens[1] * gens[2] );
26
gap> Order( ( gens[1] * gens[2] )^2 * gens[2] );
28
</Example>

<P/>

Since the <C>2D</C> elements in <M>O_7(3).2</M> lift to elements of
order four in <M>2.O_7(3).2</M>,
and since the elements of order <M>26</M> in <M>O_7(3).2</M> lift to
elements of order <M>52</M> in <M>2.O_7(3).2</M>,
the given matrices do not fit.
They would fit, however, to the isoclinic variant <M>G = 2.O_7(3).2^*</M>,
provided that they generate this group and that the first generator
is a preimage of a <C>2D</C> element (that is, <E>not</E> a preimage of a
<C>2E</C> element).

<P/>

In order to convince ourselves that the given matrices generate <M>G</M>,
we proceed as follows.
First we compute a faithful permutation representation of the group <M>H</M>,
say, that is generated by the given matrices.
Then we show that the derived subgroup <M>D</M> of <M>H</M> is a perfect group
with a central subgroup <M>Z</M> of order two such that the factor group
<M>D / Z</M> is a simple group that is isomorphic with <M>O_7(3)</M>,
thus <M>D</M> is the double cover of <M>O_7(3)</M>.
It remains to show that <M>H / Z</M> is not a direct product of <M>D / Z</M>
and a group of order two;
for that, it is enough to show that the centre of <M>H / Z</M> is trivial.

<P/>

<Example>
gap> g:= GroupWithGenerators( gens );;
gap> Size( g );
18341406720
gap> orbs:= Orbits( g, Elements( GF(3)^8 ) );;
gap> Collected( List( orbs, Length ) );
[ [ 1, 1 ], [ 2240, 1 ], [ 4320, 1 ] ]
gap> orb:= First( orbs, x -> Length( x ) = 2240 );;
gap> acthom:= ActionHomomorphism( g, orb, OnRight );;
gap> img:= Image( acthom );;
gap> Size( img ) = Size( g );
true
gap> der:= DerivedSubgroup( img );;
gap> IsPerfectGroup( der );
true
gap> z:= Centre( der );;
gap> Size( z );
2
gap> blocks:= Orbits( z, MovedPoints( img ) );;
gap> act:= Action( img, blocks, OnSets );;
gap> Size( act ) = Size( g ) / Size( z );
true
gap> IsSimple( act );
true
gap> IsomorphismTypeInfoFiniteSimpleGroup( act );
rec( name := "B(3,3) = O(7,3)", parameter := [ 3, 3 ], series := "B" )
gap> Size( Centre( act ) );
1
</Example>

<P/>

Now we show that the first generator is in fact a preimage of a
<C>2D</C> element in <M>O_7(3).2</M>;
note that this conjugacy class is the first class outside <M>O_7(3)</M>,
and it is uniquely determined by the centralizer order of its elements.

<P/>

<Example>
gap> ind:= Permutation( gens[1]^acthom, blocks, OnSets );;
gap> c:= Size( Centralizer( act, ind ) );
24261120
gap> t:= CharacterTable( "O7(3).2" );;
gap> Positions( SizesCentralizers( t ), c );
[ 53 ]
gap> Positions( OrdersClassRepresentatives( t ), 2 );
[ 2, 3, 4, 53, 54, 55 ]
</Example>

<P/>

Thus we have shown that the given matrices are standard generators
of the group <M>G</M>.

<P/>

The erroneous representation gets excluded from the &GAP; interface
by removing it from the data list in the file <F>gap/atlasprm.json</F>
of the &AtlasRep; package and then adding an entry to the global variable
<Ref Var="AGR.ExclusionList"/>.

<!-- and add entry to htm/data/changes.htm -->

<P/>

The given matrices have been added to the data that are distributed
together with the &AtlasRep; package,
the representation is now available with the name <C>2O73d2iG1-f3r8B0</C>.
(In order to make this work, also the &ATLAS; name <C>"2O73d2i"</C> for the
group with &GAP; name <C>"Isoclinic(2.O7(3).2)"</C> had to be notified
via a call to <C>AGR.GNAN</C>.

<P/>

<Example>
gap> OneAtlasGeneratingSetInfo( "Isoclinic(2.O7(3).2)", Dimension, 8,
>                               Ring, GF(3) );
rec( dim := 8, groupname := "Isoclinic(2.O7(3).2)", id := "", 
  identifier := 
    [ "Isoclinic(2.O7(3).2)", 
      [ [ "internal", "2O73d2iG1-f3r8B0.m1" ], 
          [ "internal", "2O73d2iG1-f3r8B0.m2" ] ], 1, 3 ], 
  repname := "2O73d2iG1-f3r8B0", repnr := 1, ring := GF(3), 
  standardization := 1, type := "matff" )
</Example>

<P/>

Of course we can create a representation of <M>2.O_7(3).2</M> from this
representation,
by multiplying the first generator with a fourth root of unity,
for example with <C>Z(9)^2</C>.
(see <Cite Key="CCN85" Where="p. xxiii"/>).
Note that this representation is defined over the field with <M>9</M>
elements, and that <M>2.O_7(3).2</M> does not have a faithful matrix
representation of degree <M>8</M> over the field with <M>3</M> elements.
The data that are distributed together with the &AtlasRep; package
contain also this representation, with the name <C>2O73d2G1-f9r8B0</C>.

<Example>
gap> OneAtlasGeneratingSetInfo( "2.O7(3).2", Dimension, 8, Ring, GF(9) );
rec( dim := 8, groupname := "2.O7(3).2", id := "", 
  identifier := 
    [ "2.O7(3).2", 
      [ [ "internal", "2O73d2G1-f9r8B0.m1" ], 
          [ "internal", "2O73d2G1-f9r8B0.m2" ] ], 1, 9 ], 
  repname := "2O73d2G1-f9r8B0", repnr := 1, ring := GF(3^2), 
  size := 18341406720, standardization := 1, type := "matff" )
</Example>

</Subsection>

</Section>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="subsect:F22d2_wrong_maxes">

Some of the straight line programs that are available in the web &ATLAS; for
computing generators of maximal subgroups of the group <M>G = Fi_{22}.2</M>
are not correct.
In fact, all of these programs compute generators for maximal subgroups
but for some of them, the maximal subgroup is in a class different from
the one that is claimed.

<P/>

Thus the programs in question have been excluded from the &GAP; interface
by removing them from the data list in the file <F>gap/atlasprm.json</F>
of the &AtlasRep; package and then adding entries to the global variable
<Ref Var="AGR.ExclusionList"/>.

<P/>

Then the same programs have been added to the data that are distributed
together with the &AtlasRep; package, but with different names;
the mapping of names is as follows.

<Table Align="|r|l|l|">
<HorLine>
<Row>
  <Item>Order</Item>
  <Item>Excluded name</Item><Item>Supported name</Item>
</Row>
<Row>
  <Item><M>2090188800</M></Item>
  <Item><C>F22d2G1-max12W1</C></Item><Item><C>F22d2G1-max3W2</C></Item>
</Row>
<Row>
  <Item><M>908328960</M></Item>
  <Item><C>F22d2G1-max10W1</C></Item><Item><C>F22d2G1-max4W2</C></Item>
</Row>
<Row>
  <Item><M>185794560</M></Item>
  <Item><C>F22d2G1-max3W1</C></Item><Item><C>F22d2G1-max5W2</C></Item>
</Row>
<Row>
  <Item><M>106168320</M></Item>
  <Item><C>F22d2G1-max4W1</C></Item><Item><C>F22d2G1-max6W2</C></Item>
</Row>
<Row>
  <Item><M>78382080</M></Item>
  <Item><C>F22d2G1-max5W1</C></Item><Item><C>F22d2G1-max7W2</C></Item>
</Row>
<Row>
  <Item><M>35942400</M></Item>
  <Item><C>F22d2G1-max6W1</C></Item><Item><C>F22d2G1-max8W2</C></Item>
</Row>
<Row>
  <Item><M>35389440</M></Item>
  <Item><C>F22d2G1-max7W1</C></Item><Item><C>F22d2G1-max9W2</C></Item>
</Row>
<Row>
  <Item><M>25194240</M></Item>
  <Item><C>F22d2G1-max8W1</C></Item><Item><C>F22d2G1-max10W2</C></Item>
</Row>
<Row>
  <Item><M>10077696</M></Item>
  <Item><C>F22d2G1-max9W1</C></Item><Item><C>F22d2G1-max11W2</C></Item>
</Row>
<Row>
  <Item><M>8491392</M></Item>
  <Item><C>F22d2G1-max11W1</C></Item><Item><C>F22d2G1-max12W2</C></Item>
</Row>
<HorLine>
</Table>

<P/>

(A possible reason for the different numbering could be that the classes
of maximal subgroups can be listed either according to non-increasing index
or according to the ordering in <Cite Key="CCN85" Where="p. 163"/>;
note that in the latter ordering, the relatively small <Q>novelties</Q>
<M>G_2(3):2</M> and <M>3^5:(2 \times U_4(2).2)</M> appear in the positions
<M>3</M> and <M>4</M>, respectively, whereas the positions of these groups
according to increasing index must be <M>12</M> and <M>10</M>,
respectively.
Reordering the classes from the latter ordering to the former one would
result in the mapping that appears in the above table;
unfortunately, the straight line programs had not been ordered according to
the latter ordering.).

</Subsection>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="subsect:L38d2_classreps">
<Heading>Class representatives of <M>L_3(8).2</M></Heading>


1. the script is WRONG:
   Take the available 6-dim. repres. over GF(8),
   its Brauer character value at 63AB does not fit to the char. table!

2. thus CHOOSE the available repres. as the FIRST one in the table,
   find out in which classes the inner elements lie

3. next, determine the outer classes:
   - 18A-C by their squares, which are 7th powers of the order 63 elements
   - 14G-I by their squares which are connected to the 9th powers of the
     order 63 elements
   - 8A-B by a representation

4. thus get a unique script; add it!


-> not unique program L38d2G1-cycW1 --can this be improved?
   outputs := [ "7GH", "14AB", "63AB", "73AB", "8A-B", "14G-I", "18A-C" ]
-> on 8A-B: 2r2
   on 14G-I: y7
   on 18A-C: y9
-> mod 3: 8AB disting. by dim. 72, 14G-I by 657
-> mod 7: 8AB disting. by dim. 72, 14G-I by 511
-> mod 73: 8AB disting. by dim. 71, 14G-I by 657, 18A-C by 511

-> available:  9: G ≤ GL(72,7)  -> does not help!
              23: G ≤ GL(71,73) -> two irreducibles, decidable!
              25: G ≤ GL(511a,73)
              26: G ≤ GL(511b,73)

          1a 2a 3a 4a 7a 7b 7c 7d 7e 7f 7g 9a 9b 9c 14a 14b 14c 21a 21b 21c
     2P   1a 1a 3a 2a 7b 7c 7a 7d 7f 7g 7e 9b 9c 9a  7b  7c  7a 21b 21c 21a
     3P   1a 2a 1a 4a 7c 7a 7b 7d 7g 7e 7f 3a 3a 3a 14c 14a 14b  7c  7a  7b
     7P   1a 2a 3a 4a 1a 1a 1a 1a 1a 1a 1a 9b 9c 9a  2a  2a  2a  3a  3a  3a
    73P   1a 2a 3a 4a 7c 7a 7b 7d 7g 7e 7f 9a 9b 9c 14c 14a 14b 21c 21a 21b

X.1        1  1  1  1  1  1  1  1  1  1  1  1  1  1   1   1   1   1   1   1
X.2        1  1  1  1  1  1  1  1  1  1  1  1  1  1   1   1   1   1   1   1
X.3       71  7 -1 -1  8  8  8  1  1  1  1 -1 -1 -1   .   .   .  -1  -1  -1
X.4       71  7 -1 -1  8  8  8  1  1  1  1 -1 -1 -1   .   .   .  -1  -1  -1


      2   .   .   .   .   .   .   .   .   .  4  1   5    5   1   1   1   1   1
      3   2   2   2   2   2   2   2   2   2  2  2   .    .   .   .   .   2   2
      7   1   1   1   1   1   1   1   1   1  1  .   .    .   1   1   1   .   .
     73   .   .   .   .   .   .   .   .   .  .  .   .    .   .   .   .   .   .

        63a 63b 63c 63d 63e 63f 63g 63h 63i 2b 6a  8a   8b 14d 14e 14f 18a 18b
     2P 63b 63c 63a 63e 63f 63d 63h 63i 63g 1a 3a  4a   4a  7f  7g  7e  9b  9c
     3P 21c 21a 21b 21a 21b 21c 21b 21c 21a 2b 2b  8b   8a 14f 14d 14e  6a  6a
     7P  9b  9c  9a  9a  9b  9c  9c  9a  9b 2b 6a  8a   8b  2b  2b  2b 18b 18c
    73P 63e 63f 63d 63h 63i 63g 63b 63c 63a 2b 6a  8a   8b 14f 14d 14e 18a 18b

X.1       1   1   1   1   1   1   1   1   1  1  1   1    1   1   1   1   1   1
X.2       1   1   1   1   1   1   1   1   1 -1 -1  -1   -1  -1  -1  -1  -1  -1
X.3      -1  -1  -1  -1  -1  -1  -1  -1  -1 -1 -1  AQ  *AQ  -1  -1  -1  -1  -1
X.4      -1  -1  -1  -1  -1  -1  -1  -1  -1  1  1 -AQ -*AQ   1   1   1   1   1


      2   1
      3   2
      7   .
     73   .

        18c
     2P  9a
     3P  6a
     7P 18a
    73P 18c

X.1       1
X.2      -1
X.3      -1
X.4       1


AQ = -1+2*E(8)-2*E(8)^3
  = -1+2*Sqrt(2) = -1+2r2


gap> prg:= AtlasProgram( "L3(8).2", "cyclic" );
rec( groupname := "L3(8).2", identifier := [ "L3(8).2", "L38d2G1-cycW1", 1 ],
  outputs := [ "7GH", "14AB", "63AB", "73AB", "8A-B", "14G-I", "18A-C" ],
  program := <straight line program>, standardization := 1, version := "1" )
gap> gens:= OneAtlasGeneratingSetInfo( "L3(8).2", Dimension, 71, Characteristic, 73 );
rec( dim := 71, groupname := "L3(8).2", id := "",
  identifier := [ "L3(8).2", [ "L38d2G1-f73r71B0.m1", "L38d2G1-f73r71B0.m2" ],
      1, 73 ], repname := "L38d2G1-f73r71B0", repnr := 23, ring := GF(73),
  size := 32965632, standardization := 1, type := "matff" )
gap> gens:= AtlasGenerators( gens );
rec( dim := 71,
  generators := [ < immutable compressed matrix 71x71 over GF(73) >,
      < immutable compressed matrix 71x71 over GF(73) > ],
  groupname := "L3(8).2", id := "",
  identifier := [ "L3(8).2", [ "L38d2G1-f73r71B0.m1", "L38d2G1-f73r71B0.m2" ],
      1, 73 ], repname := "L38d2G1-f73r71B0", repnr := 23, ring := GF(73),
  size := 32965632, standardization := 1, type := "matff" )
gap> res:= ResultOfStraightLineProgram( prg.program, gens.generators );;
gap> Length( res );
7
gap> 2b:= res[7]^9;
< immutable compressed matrix 71x71 over GF(73) >
gap> Order( last );
2
gap> BrauerCharacterValue( 2b );
-1
gap> # thus we have 71a!
gap> 8ab:= res[5];
< immutable compressed matrix 71x71 over GF(73) >
gap> Order( last );
8
gap> BrauerCharacterValue( 8ab );
-1-2*E(8)+2*E(8)^3
gap> Quadratic( last );
rec( ATLAS := "-1-2r2", a := -1, b := -2, d := 1, display := "-1-2*Sqrt(2)",
  root := 2 )
gap> # thus we have class 8B!


concerning 18A-C:
squares are in 9B, 9C, 9A
and 9A is the 7th power of 63c, 63d, 63h
and 9B is the 7th power of 63a, 63e, 63i
and 9C is the 7th power of 63b, 63f, 63g

-> so we could solve this without a repres.! (and the answer defines some repres.)

-> what about 14G-I?

squares:  14G -> 7f
          14H -> 7g
          14I -> 7e

and 9th powers of 63 are
     63a -> 21c -> 7b
     63b -> 21a -> 7c
     63c -> 21b -> 7a
     63d -> 21a -> 7c
     63e -> 21b -> 7a
     63f -> 21c -> 7b
     63g -> 21b -> 7a
     63h -> 21c -> 7b
     63i -> 21a -> 7c

and there are representations of degree 146 or 1168 or 657 which couple 7a-c to 7d-f

23: G ≤ GL(71,73)
24: G ≤ GL(441,73)
25: G ≤ GL(511a,73)
26: G ≤ GL(511b,73)

-> but we do not have it!

-> in char. 3, also candidates:
   degrees 146, 657

-> use the nat. repres. in char. 2!



</Subsection>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Section Label="sect:L34_maxes">
<Heading>Maximal Subgroups of <M>L_3(4)</M></Heading>

Two straight line programs that are available in the web &ATLAS; for
computing generators of maximal subgroups of the group <M>G = L_3(4)</M>
are not correct.

<P/>

The contents of the file <C>L34G1-max4W1</C> is as follows.

<P/>

<Listing>
mu 1 2 3
mu 2 1 4
iv 3 5
mu 4 4 2
mu 2 5 4
mu 4 3 2
</Listing>

<P/>

However,
if we apply this program to standard generators of <M>G</M> then
the outputs generate the whole group.

<P/>

<Example>
gap> slp:= ScanStraightLineProgram( "\
> mu 1 2 3\n\
> mu 2 1 4\n\
> iv 3 5\n\
> mu 4 4 2\n\
> mu 2 5 4\n\
> mu 4 3 2", "string" );
rec( program := <straight line program> )
gap> g:= AtlasGroup( "L3(4)" );
Group([ (1,2)(4,6)(5,7)(8,12)(9,14)(10,15)(11,17)(13,19), 
  (2,3,5,4)(6,8,13,9)(7,10,16,11)(12,18)(14,20,21,15)(17,19) ])
gap> res:= ResultOfStraightLineProgram( slp.program,
>              GeneratorsOfGroup( g ) );;
gap> Size( SubgroupNC( g, res ) );
20160
</Example>

<P/>

Similarly, the file <C>L34G1-max5W1</C> has the following contents.

<P/>

<Listing>
mu 1 2 3
mu 2 1 4
iv 3 5
mu 4 5 2
mu 2 3 4
mu 4 3 2
</Listing>

<P/>

The subgroup generated by the outputs of this program has order <M>10</M>
and is hence too small.

<P/>

<Example>
gap> slp:= ScanStraightLineProgram( "\
> mu 1 2 3\n\
> mu 2 1 4\n\
> iv 3 5\n\
> mu 4 5 2\n\
> mu 2 3 4\n\
> mu 4 3 2", "string" );
rec( program := <straight line program> )
gap> res:= ResultOfStraightLineProgram( slp.program,
>              GeneratorsOfGroup( g ) );;
gap> Size( SubgroupNC( g, res ) );
10
</Example>

<P/>

Now we want to replace the wrong programs by correct ones.
According to <Cite Key="CCN85" Where="p. 23"/>,
the subgroups in the 3rd, 4th, and 5th class of maximal subgroups of <M>G</M>
are all isomorphic with the alternating group <M>A_6</M>.
Thus our task is to find two subgroups of type <M>A_6</M> in <M>G</M>
that are not conjugate to each other and also not conjugate to the
representative of the 3rd class of maximal subgroups.

<P/>

Our approach is to keep the first generator of <M>G</M> (an involution),
and to iterate over short words in a free monoid until the corresponding
word in the standard generators of <M>G</M> together with the involution
generate a suitable subgroup.
(Note that all subgroups of order <M>360</M> in <M>G</M> are maximal in
<M>G</M> and have the type <M>A_6</M>.)

<P/>

<Example>
gap> g:= AtlasGroup( "L3(4)" );;
gap> s3:= AtlasSubgroup( "L3(4)", 3 );;
gap> Size( s3 );  IsSimple( s3 );
360
true
gap> gens:= ShallowCopy( GeneratorsOfGroup( g ) );;
gap> f:= FreeMonoid( 2 );;
gap> fgens:= GeneratorsOfMonoid( f );;
[ m1, m2 ]
gap> iter:= Iterator( f );;
gap> repeat
> w4:= NextIterator( iter );
> s4:= Group( gens[1], MappedWord( w4, fgens, gens ) );
> until Size( s4 ) = 360 and not IsConjugate( g, s3, s4 );
gap> w4;
m2*m1*m2^3*m1*m2*m1*m2
gap> repeat
> w5:= NextIterator( iter );
> s5:= Group( gens[1], MappedWord( w5, fgens, gens ) );
> until Size( s5 ) = 360
> and not IsConjugate( g, s3, s5 )
> and not IsConjugate( g, s4, s5 );
gap> w5;
m2*m1*m2*m1*m2^3*m1*m2
</Example>

<P/>

The erroneous programs get excluded from the &GAP; interface
by removing them from the data list in the file <F>gap/atlasprm.json</F>
of the &AtlasRep; package and then adding entries to the global variable
<Ref Var="AGR.ExclusionList"/>.

<P/>

The above words have been turned into straight line programs and then added
to the data that are distributed together with the &AtlasRep; package,
the programs are now available with the names <C>L34G1-max4W2</C>;
and <C>L34G1-max5W2</C>, respectively;
note that the names <C>L34G1-max4W1</C> and <C>L34G1-max5W1</C>
must be avoided.

<!-- and add entries to htm/data/changes.htm -->

<P/>

<Example>
gap> subs:= List( [3..5], i -> AtlasSubgroup( g, i ) );;
gap> List( subs, Size );
[ 360, 360, 360 ]
gap> IsConjugate( g, subs[1], subs[2] );
false
gap> IsConjugate( g, subs[1], subs[3] );
false
gap> IsConjugate( g, subs[2], subs[3] );
false
</Example>

<P/>

It turns out that the generators of the subgroups are in fact
standard generators of <M>A_6</M>.

<P/>

<Example>
gap> prg:= AtlasProgram( "A6", 1, "check" );
rec( groupname := "A6", identifier := [ "A6", "A6G1-check1", 1, 1 ], 
  program := <straight line decision>, standardization := 1, version := "1" )
gap> ForAll( subs, s -> ResultOfStraightLineDecision( prg.program,
>                           GeneratorsOfGroup( s ) ) );
true
</Example>

<P/>

Thus we can provide also the (empty) straight line programs
<C>L34G1max3W1-A6G1W1</C>, <C>L34G1max4W2-A6G1W1</C>, and
<C>L34G1max5W2-A6G1W1</C>,
which express that the &GAP; interface provides <E>standard</E> generators
for the maximal subgroups in question.

</Section>


<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->

</Chapter>