examples.xml

<!-- #################################################################### -->
<!-- ##                                                                ## -->
<!-- ##  examples.xml         RCWA documentation          Stefan Kohl  ## -->
<!-- ##                                                                ## -->
<!-- #################################################################### -->

<Chapter Label="ch:Examples"><Heading>Examples</Heading>

<Ignore Remark="settings for the example tester">
<Example>
<![CDATA[
gap> SizeScreen([75,24]);;
gap> SetAssertionLevel(0);
]]>
</Example>
</Ignore>

This chapter discusses a number of examples of rcwa mappings
and -groups in detail. All of them show different aspects of the package,
and the order in which they appear is entirely arbitrary. In particular
they are not ordered by degree of difficulty or interest. <P/>

<Index Key="LoadRCWAExamples"><C>LoadRCWAExamples</C></Index>
<Index Key="AssignGlobals"><C>AssignGlobals</C></Index>

The rcwa mappings, rcwa groups and other objects defined in this chapter can
be found in the file <F>pkg/rcwa/examples/examples.g</F>. This file can be
read into the current &GAP; session by the function
<Ref Func="LoadRCWAExamples"/> which takes no arguments and returns the name
of a variable which the record containing the examples got assigned to.
The global variable assignments made in a section of this chapter can be
made by applying the function <C>AssignGlobals</C> to the respective
component of the examples record. The component names are given at the end
of the corresponding sections. <P/>

The discussions of the examples are typically far from being exhaustive.
It is quite likely that in many instances by just a few little modifications
or additional easy commands you can find out interesting things yourself
-- have fun!

<!-- #################################################################### -->

<Section Label="sec:ThompsonsGroupV">
<Heading>
  Thompson's group V
</Heading>

Thompson's group V, also known as Higman-Thompson group, is a finitely
presented infinite simple group. This group has been found by Graham Higman,
cf.&nbsp;<Cite Key="Higman74"/>. We show that the group

<Example>
<![CDATA[
gap> G := Group(List([[0,2,1,4],[0,4,1,4],[1,4,2,4],[2,4,3,4]],
>                    ClassTransposition));
<(0(2),1(4)),(0(4),1(4)),(1(4),2(4)),(2(4),3(4))>
]]>
</Example>

is isomorphic to Thompson's group V. 
This isomorphism has been pointed out by John P. McDermott.
We take a slightly different set of generators:

<Example>
<![CDATA[
gap> k := ClassTransposition(0,2,1,2);;
gap> l := ClassTransposition(1,2,2,4);;
gap> m := ClassTransposition(0,2,1,4);;
gap> n := ClassTransposition(1,4,2,4);;
gap> H := Group(k,l,m,n);
<(0(2),1(2)),(1(2),2(4)),(0(2),1(4)),(1(4),2(4))>
gap> G = H; # k, l, m and n generate G as well
true
]]>
</Example>

Now we verify that our four generators satisfy the relations given on
page&nbsp;50 in&nbsp;<Cite Key="Higman74"/>, when we read <C>k</C>
as&nbsp;<M>\kappa</M>, <C>l</C> as&nbsp;<M>\lambda</M>, <C>m</C>
as&nbsp;<M>\mu</M> and <C>n</C> as&nbsp;<M>\nu</M>:

<Example>
<![CDATA[
gap> HigmanThompsonRels :=
> [ k^2, l^2, m^2, n^2,                           # (1) in Higman's book
>   l*k*m*k*l*n*k*n*m*k*l*k*m,                    # (2)        "
>   k*n*l*k*m*n*k*l*n*m*n*l*n*m,                  # (3)        "
>   (l*k*m*k*l*n)^3, (m*k*l*k*m*n)^3,             # (4)        "
>   (l*n*m)^2*k*(m*n*l)^2*k,                      # (5)        "
>   (l*n*m*n)^5,                                  # (6)        "
>   (l*k*n*k*l*n)^3*k*n*k*(m*k*n*k*m*n)^3*k*n*k*n,# (7)        "
>   ((l*k*m*n)^2*(m*k*l*n)^2)^3,                  # (8)        "
>   (l*n*l*k*m*k*m*n*l*n*m*k*m*k)^4,              # (9)        "
>   (m*n*m*k*l*k*l*n*m*n*l*k*l*k)^4,              #(10)        "
>   (l*m*k*l*k*m*l*k*n*k)^2,                      #(11)        "
>   (m*l*k*m*k*l*m*k*n*k)^2 ];                    #(12)        "
[ IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ), 
  IdentityMapping( Integers ), IdentityMapping( Integers ) ]
]]>
</Example>

We conclude that our group is an homomorphic image of Thompson's group V.
But since Thompson's group V is simple and our group is not trivial,
this means indeed that the two groups are isomorphic. <P/>

In fact it is straightforward to show that <C>G</C> is the group
<C>CT([2],Integers)</C> which is generated by the set of all
class transpositions which interchange residue classes modulo powers
of&nbsp;2. First we check that <C>G</C> contains all 11 class
transpositions which interchange residue classes modulo 2 or&nbsp;4:

<Example>
<![CDATA[
gap> S := Filtered(List(ClassPairs(4),ClassTransposition),
>                  ct->Mod(ct) in [2,4]);
[ ( 0(2), 1(2) ), ( 0(2), 1(4) ), ( 0(2), 3(4) ), ( 0(4), 1(4) ), 
  ( 0(4), 2(4) ), ( 0(4), 3(4) ), ( 1(2), 0(4) ), ( 1(2), 2(4) ), 
  ( 1(4), 2(4) ), ( 1(4), 3(4) ), ( 2(4), 3(4) ) ]
gap> IsSubset(G,S);
true
]]>
</Example>

Then we give a function which takes a class transposition
<M>\tau \in {\rm CT}_\emptyset(&ZZ;)</M>, and which returns
a factorization of an element <M>\gamma</M> satisfying
<M>\tau^\gamma \in S</M> into <M>g_1 := \tau_{0(2),1(4)} \in S</M>,
<M>g_2 := \tau_{0(2),3(4)} \in S</M>, <M>g_3 := \tau_{1(2),0(4)} \in S</M>,
<M>g_4 := \tau_{1(2),2(4)} \in S</M>, <M>h_1 := \tau_{0(4),1(4)} \in S</M>
and <M>h_2 := \tau_{1(4),2(4)} \in S</M>:

<Listing Type="GAP code">
<![CDATA[
ReducingConjugator := function ( tau )

  local  w, F, g1, g2, g3, g4, h1, h2, h, cls, cl, r;

  g1 := ClassTransposition(0,2,1,4); h1 := ClassTransposition(0,4,1,4);
  g2 := ClassTransposition(0,2,3,4); h2 := ClassTransposition(1,4,2,4);
  g3 := ClassTransposition(1,2,0,4);
  g4 := ClassTransposition(1,2,2,4);

  F := FreeGroup("g1","g2","g3","g4","h1","h2");

  w := One(F); if Mod(tau) <= 4 then return w; fi;

  # Before we can reduce the moduli of the interchanged residue classes,
  # we must make sure that both of them have at least modulus 4.
  cls := TransposedClasses(tau);
  if Mod(cls[1]) = 2 then
    if Residue(cls[1]) = 0 then
      if Residue(cls[2]) mod 4 = 1 then tau := tau^g2; w := w * F.2;
                                   else tau := tau^g1; w := w * F.1; fi;
    else
      if Residue(cls[2]) mod 4 = 0 then tau := tau^g4; w := w * F.4;
                                   else tau := tau^g3; w := w * F.3; fi;
    fi;
  fi;

  while Mod(tau) > 4 do # Now we can successively reduce the moduli.
    if not ForAny(AllResidueClassesModulo(2),
                  cl -> IsEmpty(Intersection(cl,Support(tau))))
    then
      cls := TransposedClasses(tau);
      h := Filtered([h1,h2],
             hi->Length(Filtered(cls,cl->IsSubset(Support(hi),cl)))=1);
      h := h[1]; tau := tau^h;
      if h = h1 then w := w * F.5; else w := w * F.6; fi;
    fi;
    cl := TransposedClasses(tau)[2]; # class with larger modulus
    r  := Residue(cl);
    if   r mod 4 = 1 then tau := tau^g1; w := w * F.1;
    elif r mod 4 = 3 then tau := tau^g2; w := w * F.2;
    elif r mod 4 = 0 then tau := tau^g3; w := w * F.3;
    elif r mod 4 = 2 then tau := tau^g4; w := w * F.4; fi;
  od;

  return w;
end;
]]>
</Listing>

After assigning <C>g1</C>, <C>g2</C>, <C>g3</C>, <C>g4</C>, <C>h1</C>
and&nbsp;<C>h2</C> appropriately, we obtain for example:

<Listing Type="Example">
<![CDATA[
gap> ReducingConjugator(ClassTransposition(3,16,34,256));
h2*g1*h1*g1*h1*g1*h1*g1*h2*g2*h2*g4*h2*g4*h2*g3
gap> gamma := h2*g1*h1*g1*h1*g1*h1*g1*h2*g2*h2*g4*h2*g4*h2*g3;
<rcwa permutation of Z with modulus 256>
gap> ct := ClassTransposition(3,16,34,256)^gamma;;
gap> IsClassTransposition(ct);;
gap> ct;
ClassTransposition(1,4,2,4)
]]>
</Listing>

Thompson's group V can also be embedded in a natural way into CT(GF(2)[x]):

<Example>
<![CDATA[
gap> x := Indeterminate(GF(2));; SetName(x,"x");
gap> R := PolynomialRing(GF(2),1);;
gap> k := ClassTransposition(0,x,1,x);;
gap> l := ClassTransposition(1,x,x,x^2);;
gap> m := ClassTransposition(0,x,1,x^2);;
gap> n := ClassTransposition(1,x^2,x,x^2);;
gap> G := Group(k,l,m,n);
<rcwa group over GF(2)[x] with 4 generators>
]]>
</Example>

The correctness of this representation can likewise be verified by simply
checking the defining relations given above. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().HigmanThompson);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:FactoringTheCollatzPermutation">
<Heading>
  Factoring Collatz' permutation of the integers
</Heading>

In 1932, Lothar Collatz mentioned in his notebook the following permutation
of the integers:

<Example>
<![CDATA[
gap> Collatz := RcwaMapping([[2,0,3],[4,-1,3],[4,1,3]]);;
gap> Display(Collatz);

Rcwa mapping of Z with modulus 3

        /
        | 2n/3     if n in 0(3)
 n |-> <  (4n-1)/3 if n in 1(3)
        | (4n+1)/3 if n in 2(3)
        \

gap> ShortCycles(Collatz,[-50..50],50); # There are some finite cycles:
[ [ 0 ], [ -1 ], [ 1 ], [ 2, 3 ], [ -2, -3 ], [ 4, 5, 7, 9, 6 ], 
  [ -4, -5, -7, -9, -6 ], 
  [ 44, 59, 79, 105, 70, 93, 62, 83, 111, 74, 99, 66 ], 
  [ -44, -59, -79, -105, -70, -93, -62, -83, -111, -74, -99, -66 ] ]
]]>
</Example>

The cycle structure of Collatz' permutation has not been completely
determined yet. In particular it is not known whether the cycle
containing&nbsp;8 is finite or infinite. 
Nevertheless, the factorization routine included in this package can
determine a factorization of this permutation into class transpositions,
i.e. involutions interchanging two disjoint residue classes:

<Example>
<![CDATA[
gap> Collatz in CT(Integers);  # `Collatz' lies in the simple group CT(Z).
true
gap> Length(Factorization(Collatz));
212
]]>
</Example>

Setting the Info level of <C>InfoRCWA</C> equal to&nbsp;2 (simply issue
<C>RCWAInfo(2);</C>) causes the factorization routine to display detailed
information on the progress of the factoring process. For reasons of saving
space, this is not done in this manual. <P/>

We would like to get a factorization into fewer factors. Firstly, we try
to factor the inverse -- just like the various options interpreted by the
factorization routine, this has influence on decisions taken during the
factoring process:

<Example>
<![CDATA[
gap> Length(Factorization(Collatz^-1));
129
]]>
</Example>

This is already a shorter product, but can still be improved.
We remember the <C>mKnot</C>'s, of which the permutation <C>mKnot(3)</C>
looks very similar to Collatz' permutation. Therefore it is straightforward
to try to factor both <C>mKnot(3)</C> and <C>Collatz/mKnot(3)</C>, and to
look whether the sum of the numbers of factors is less than&nbsp;129:

<Example>
<![CDATA[
gap> KnotFacts := Factorization(mKnot(3));;
gap> QuotFacts := Factorization(Collatz/mKnot(3));;
gap> List([KnotFacts,QuotFacts],Length);
[ 59, 9 ]
gap> CollatzFacts := Concatenation(QuotFacts,KnotFacts);
[ ( 0(6), 4(6) ), ( 0(6), 5(6) ), ( 0(6), 3(6) ), ( 0(6), 1(6) ), 
  ( 0(6), 2(6) ), ( 2(3), 4(6) ), ( 0(3), 4(6) ), ( 2(3), 1(6) ), 
  ( 0(3), 1(6) ), ( 0(36), 35(36) ), ( 0(36), 22(36) ), 
  ( 0(36), 18(36) ), ( 0(36), 17(36) ), ( 0(36), 14(36) ), 
  ( 0(36), 20(36) ), ( 0(36), 4(36) ), ( 2(36), 8(36) ), 
  ( 2(36), 16(36) ), ( 2(36), 13(36) ), ( 2(36), 9(36) ), 
  ( 2(36), 7(36) ), ( 2(36), 6(36) ), ( 2(36), 3(36) ), 
  ( 2(36), 10(36) ), ( 2(36), 15(36) ), ( 2(36), 12(36) ), 
  ( 2(36), 5(36) ), ( 21(36), 28(36) ), ( 21(36), 33(36) ), 
  ( 21(36), 30(36) ), ( 21(36), 23(36) ), ( 21(36), 34(36) ), 
  ( 21(36), 31(36) ), ( 21(36), 27(36) ), ( 21(36), 25(36) ), 
  ( 21(36), 24(36) ), ( 26(36), 32(36) ), ( 26(36), 29(36) ), 
  ( 10(18), 35(36) ), ( 5(18), 35(36) ), ( 10(18), 17(36) ), 
  ( 5(18), 17(36) ), ( 8(12), 14(24) ), ( 6(9), 17(18) ), 
  ( 3(9), 17(18) ), ( 0(9), 17(18) ), ( 6(9), 16(18) ), ( 3(9), 16(18) ),
  ( 0(9), 16(18) ), ( 6(9), 11(18) ), ( 3(9), 11(18) ), ( 0(9), 11(18) ),
  ( 6(9), 4(18) ), ( 3(9), 4(18) ), ( 0(9), 4(18) ), ( 0(6), 14(24) ), 
  ( 0(6), 2(24) ), ( 8(12), 17(18) ), ( 7(12), 17(18) ), 
  ( 8(12), 11(18) ), ( 7(12), 11(18) ), PrimeSwitch(3)^-1, 
  ( 7(12), 17(18) ), ( 2(6), 17(18) ), ( 0(3), 17(18) ), 
  PrimeSwitch(3)^-1, PrimeSwitch(3)^-1, PrimeSwitch(3)^-1 ]
gap> Product(CollatzFacts) = Collatz; # Check.
true
]]>
</Example>

The factors <C>PrimeSwitch(3)</C> are products of 6 class transpositions
(cf.&nbsp;<Ref Func="PrimeSwitch" Label="p"/>). <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzlikePerms);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:3nplus1Group">
<Heading>
  The <M>3n+1</M> group
</Heading>

The following group acts transitively on the set of positive integers
for which the <M>3n+1</M> conjecture holds and which are not divisible by 6:

<Example>
<![CDATA[
gap> a := ClassTransposition(1,2,4,6);;
gap> b := ClassTransposition(1,3,2,6);;
gap> c := ClassTransposition(2,3,4,6);;
gap> G := Group(a,b,c);
<(1(2),4(6)),(1(3),2(6)),(2(3),4(6))>
gap> LoadDatabaseOfGroupsGeneratedBy3ClassTranspositions();
"3CTsGroups6"
gap> 3CTsGroups6.Id3CTsGroup(G,3CTsGroups6.grps); # 'catalogue number' of G
44132
]]>
</Example>

To see this, consider the action of <M>G</M> on the <Q><M>3n+1</M> tree</Q>.
The vertices of this tree are the positive integers for which the <M>3n+1</M>
conjecture holds, and for every vertex <M>n</M> there is an edge from
<M>n</M> to <M>T(n)</M>, where <M>T</M> denotes the Collatz mapping 
<Alt Only="LaTeX">
  <Display>
  <![CDATA[T: \ \ \mathbb{Z} \longrightarrow \mathbb{Z}, \ \ \ \
    n \ \longmapsto \
    \begin{cases}
      \frac{n}{2}    & \text{if} \ \ n \ \ \text{is even}, \\
      \frac{3n+1}{2} & \text{if} \ \ n \ \ \text{is odd}
    \end{cases}]]>
  </Display>
</Alt>
<Alt Only="HTML"><![CDATA[<center>
  <img src = "collatz.png" width = "342" height = "63"
       alt = "T: Z -> Z, n |-> (n/2 if n even, (3n+1)/2 if n odd)"/>
</center>]]></Alt>
<Alt Only="Text"><Verb><![CDATA[
                                      /
                                      | n/2 if n even,
               T:  Z -> Z,   n  |->  <
                                      | (3n+1)/2 if n odd
                                      \
]]></Verb></Alt>
(cf. Chapter&nbsp;<Ref Label="ch:AboutRCWA"/>).
It is easy to check that for every vertex <M>n</M>, either <M>a</M>,
<M>b</M> or <M>c</M> maps <M>n</M> to <M>T(n)</M>, and that the
other two generators either fix <M>n</M> or map it to one of its
preimages under&nbsp;<M>T</M>.
So the <M>3n+1</M> conjecture is equivalent to the assertion that the
group <M>G</M> acts transitively on <M>\mathbb{N} \setminus 0(6)</M>.
First let's have a look at balls of small radius about 1 under the
action of <M>G</M> -- these consist of those numbers whose trajectory
under <M>T</M> reaches 1 quickly: 

<Example>
<![CDATA[
gap> Ball(G,1,5,OnPoints);
[ 1, 2, 4, 5, 8, 10, 16, 32, 64 ]
gap> Ball(G,1,10,OnPoints);
[ 1, 2, 3, 4, 5, 8, 10, 13, 16, 20, 21, 26, 32, 40, 52, 53, 64, 80, 85, 
  128, 160, 170, 256, 320, 340, 341, 512, 1024, 2048 ]
gap> Ball(G,1,15,OnPoints);
[ 1, 2, 3, 4, 5, 7, 8, 10, 11, 13, 16, 17, 20, 21, 22, 23, 26, 32, 34, 
  35, 40, 44, 45, 46, 52, 53, 64, 68, 69, 70, 75, 80, 85, 104, 106, 113, 
  128, 136, 140, 141, 151, 160, 170, 208, 212, 213, 226, 227, 256, 272, 
  277, 280, 301, 302, 320, 340, 341, 416, 424, 452, 453, 454, 512, 640, 
  680, 682, 832, 848, 853, 904, 908, 909, 1024, 1280, 1360, 1364, 1365, 
  1664, 1696, 1706, 1808, 1813, 1816, 2048, 2560, 2720, 2728, 4096, 
  5120, 5440, 5456, 5461, 8192, 10240, 10880, 10912, 10922, 16384, 
  32768, 65536 ]
gap> Ball(G,1,15,OnPoints:Spheres);
[ [ 1 ], [ 2, 4 ], [ 8 ], [ 16 ], [ 5, 32 ], [ 10, 64 ], 
  [ 3, 20, 21, 128 ], [ 40, 256 ], [ 13, 80, 85, 512 ], 
  [ 26, 160, 170, 1024 ], [ 52, 53, 320, 340, 341, 2048 ], 
  [ 17, 104, 106, 113, 640, 680, 682, 4096 ], 
  [ 34, 35, 208, 212, 213, 226, 227, 1280, 1360, 1364, 1365, 8192 ], 
  [ 11, 68, 69, 70, 75, 416, 424, 452, 453, 454, 2560, 2720, 2728, 16384 
     ], 
  [ 22, 23, 136, 140, 141, 151, 832, 848, 853, 904, 908, 909, 5120, 
      5440, 5456, 5461, 32768 ], 
  [ 7, 44, 45, 46, 272, 277, 280, 301, 302, 1664, 1696, 1706, 1808, 
      1813, 1816, 10240, 10880, 10912, 10922, 65536 ] ]
gap> List(Ball(G,1,50,OnPoints:Spheres),Length);
[ 1, 2, 1, 1, 2, 2, 4, 2, 4, 4, 6, 8, 12, 14, 17, 20, 26, 32, 43, 52, 
  66, 81, 104, 133, 170, 211, 271, 335, 424, 542, 686, 873, 1096, 1376, 
  1730, 2205, 2794, 3522, 4429, 5611, 7100, 8978, 11343, 14296, 18058, 
  22828, 28924, 36532, 46146, 58399, 73713 ]
gap> FloatQuotientsList(last);
[ 2., 0.5, 1., 2., 1., 2., 0.5, 2., 1., 1.5, 1.33333, 1.5, 1.16667, 
  1.21429, 1.17647, 1.3, 1.23077, 1.34375, 1.2093, 1.26923, 1.22727, 
  1.28395, 1.27885, 1.2782, 1.24118, 1.28436, 1.23616, 1.26567, 1.2783, 
  1.26568, 1.27259, 1.25544, 1.25547, 1.25727, 1.27457, 1.26712, 
  1.26056, 1.25752, 1.26688, 1.26537, 1.26451, 1.26342, 1.26034, 
  1.26315, 1.26415, 1.26704, 1.26303, 1.26317, 1.26553, 1.26223 ]
gap> Difference(Filtered([1..100],n->n mod 6 <> 0),Ball(G,1,40,OnPoints));
[ 27, 31, 41, 47, 55, 62, 63, 71, 73, 82, 83, 91, 94, 95, 97 ]
gap> T := RcwaMapping([[1,0,2],[3,1,2]]);;
gap> List(last2,n->Length(Trajectory(T,n,[1])));
[ 71, 68, 70, 67, 72, 69, 69, 66, 74, 71, 71, 60, 68, 68, 76 ]
]]>
</Example>

It is convenient to define an epimorphism from the free group of rank 3
to <M>G</M>:

<Example>
<![CDATA[
gap> F := FreeGroup("a","b","c");
<free group on the generators [ a, b, c ]>
gap> phi := EpimorphismByGenerators(F,G);
[ a, b, c ] -> [ ( 1(2), 4(6) ), ( 1(3), 2(6) ), ( 2(3), 4(6) ) ]
]]>
</Example>

We can compute balls about 1 in <M>G</M>:

<Example>
<![CDATA[
gap> B := Ball(G,One(G),7:Spheres);;
gap> List(B,Length);
[ 1, 3, 6, 12, 24, 48, 96, 192 ]
gap> List(B[3],Order);
[ 12, infinity, infinity, infinity, infinity, 12 ]
gap> List(B[3],g->PreImagesRepresentative(phi,g));
[ b*a, c*b, c*a, b*c, a*c, a*b ]
gap> g := a*b;; Order(g);;
gap> Display(g);

Rcwa permutation of Z with modulus 18, of order 12

( 1(6), 8(36), 4(18), 2(12) ) ( 3(6), 20(36), 10(18) )
( 5(6), 32(36), 16(18) )

]]>
</Example>

Spending some more time to compute <C>B := Ball(G,One(G),12:Spheres);;</C>,
one can check that <M>(ab)^{12}</M> is the shortest word in the generators
of <M>G</M> which does not represent the identity in the free product of
3 cyclic groups of order 2, but which represents the identity in <M>G</M>.
However, the group <M>G</M> has elements of other finite orders as well --
for example:

<Example>
<![CDATA[
gap> g := (b*a)^3*b*c;; Order(g);;
gap> Display(g);

Rcwa permutation of Z with modulus 36, of order 105

( 8(9), 16(18), 64(72), 256(288), 85(96), 128(144), 32(36) )
( 7(12), 11(18), 22(36) ) ( 5(18), 10(36), 40(144), 13(48), 
  20(72) ) ( 1(24), 2(36), 4(72) ) ( 14(36), 28(72), 112(288), 
  37(96), 56(144) )

gap> Order(a*c*b*a*b*c*a*c);
60
]]>
</Example>

With some more efforts, one finds that e.g. <M>(abc)^2c^b</M> has order
616, that <M>(abc)^2b</M> has order 2310, that <M>(ab)^2a^ca^bc</M> has
order 27720, and that <M>a(c(ab)^2)^2</M> has order 65520.
Of course <M>G</M> has many elements of infinite order as well.
Some of them have infinite cycles, like e.g.

<Example>
<![CDATA[
gap> g := b*c;;
gap> Display(g);

Rcwa permutation of Z with modulus 12

        /
        | 4n  if n in 1(3)
        | 2n  if n in 5(6)
 n |-> <  n/2 if n in 2(12)
        | n/4 if n in 8(12)
        | n   if n in 0(3)
        \

gap> Sinks(g);
[ 4(12) ]
gap> Trajectory(g,last[1],10);
[ 4(12), 16(48), 64(192), 256(768), 1024(3072), 4096(12288), 
  16384(49152), 65536(196608), 262144(786432), 1048576(3145728) ]
gap> Trajectory(g,4,20);
[ 4, 16, 64, 256, 1024, 4096, 16384, 65536, 262144, 1048576, 4194304, 
  16777216, 67108864, 268435456, 1073741824, 4294967296, 17179869184, 
  68719476736, 274877906944, 1099511627776 ]
]]>
</Example>

Others seem to have only finite cycles. Some of these appear to have
<Q>on average</Q> comparatively <Q>short</Q> cycles, like e.g.
 
<Example>
<![CDATA[
gap> g := a*b*a*c*b*c;
<rcwa permutation of Z with modulus 144>
gap> cycs := ShortCycles(g,[0..10000],100,10^20);;
gap> Difference([0..10000],Union(cycs));
[  ]
gap> Collected(List(cycs,Length));
[ [ 1, 2222 ], [ 3, 1945 ], [ 4, 1111 ], [ 5, 93 ], [ 6, 926 ], 
  [ 7, 31 ], [ 8, 864 ], [ 9, 10 ], [ 10, 289 ], [ 11, 4 ], [ 12, 95 ], 
  [ 13, 1 ], [ 14, 31 ], [ 16, 12 ], [ 18, 4 ], [ 20, 1 ] ]
]]>
</Example>

If the cycle of <M>g</M> containing some <M>n \in &ZZ;</M> is finite
and has a certain length <M>l</M>, then there is some <M>m \in &ZZ;</M>
such that for every <M>k \in &ZZ;</M> the cycle of <M>g</M> containing
<M>n + km</M> has length <M>l</M> as well. Thus, in other words, every
finite cycle of <M>g</M> <Q>belongs to</Q> a cycle of residue classes.
(This is a special property of <M>g</M> which is not shared by every
rcwa permutation -- cf. e.g. Collatz' permutation from
Section&nbsp;<Ref Label="sec:FactoringTheCollatzPermutation"/>.)
We can find some of these infinitely many <Q>residue class cycles</Q>:

<Example>
<![CDATA[
gap> cycsrc := ShortResidueClassCycles(g,Mod(g),20);
[ [ 0(6) ], [ 3(6), 160(288), 20(36) ], 
  [ 7(18), 352(864), 44(108), 28(72) ], 
  [ 11(18), 544(864), 2896(4608), 362(576), 68(108), 88(144) ], 
  [ 13(18), 640(864), 80(108), 52(72) ], [ 10(36) ], [ 34(36) ], 
  [ 1(54), 64(2592), 8(324), 4(216), 16(1152), 2(144) ], 
  [ 5(54), 256(2592), 1360(13824), 170(1728), 32(324), 40(432), 
      208(2304), 26(288) ], 
  [ 17(54), 832(2592), 4432(13824), 23632(73728), 2954(9216), 554(1728), 
      104(324), 136(432) ], 
  [ 37(54), 1792(2592), 224(324), 148(216), 784(1152), 98(144) ], 
  [ 41(54), 1984(2592), 10576(13824), 1322(1728), 248(324), 328(432), 
      1744(2304), 218(288) ], 
  [ 53(54), 2560(2592), 13648(13824), 72784(73728), 9098(9216), 
      1706(1728), 320(324), 424(432) ], [ 38(72), 58(108), 304(576) ], 
  [ 62(72), 94(108), 496(576) ] ]
gap> List(cycsrc,Length);
[ 1, 3, 4, 6, 4, 1, 1, 6, 8, 8, 6, 8, 8, 3, 3 ]
gap> Sum(List(Flat(cycsrc),cl->1/Mod(cl)));
97459/110592
gap> Float(last); # about 88% 'coverage'
0.881248
gap> cycsrc := ShortResidueClassCycles(g,3*Mod(g),20);
[ [ 0(6) ], [ 3(6), 160(288), 20(36) ], 
  [ 7(18), 352(864), 44(108), 28(72) ], 
  [ 11(18), 544(864), 2896(4608), 362(576), 68(108), 88(144) ], 
  [ 13(18), 640(864), 80(108), 52(72) ], [ 10(36) ], [ 34(36) ], 
  [ 1(54), 64(2592), 8(324), 4(216), 16(1152), 2(144) ], 
  [ 5(54), 256(2592), 1360(13824), 170(1728), 32(324), 40(432), 
      208(2304), 26(288) ], 
  [ 17(54), 832(2592), 4432(13824), 23632(73728), 2954(9216), 554(1728), 
      104(324), 136(432) ], 
  [ 37(54), 1792(2592), 224(324), 148(216), 784(1152), 98(144) ], 
  [ 41(54), 1984(2592), 10576(13824), 1322(1728), 248(324), 328(432), 
      1744(2304), 218(288) ], 
  [ 53(54), 2560(2592), 13648(13824), 72784(73728), 9098(9216), 
      1706(1728), 320(324), 424(432) ], [ 38(72), 58(108), 304(576) ], 
  [ 62(72), 94(108), 496(576) ], 
  [ 23(162), 1120(7776), 5968(41472), 746(5184), 140(972), 184(1296), 
      976(6912), 5200(36864), 650(4608), 122(864) ], 
  [ 35(162), 1696(7776), 9040(41472), 48208(221184), 257104(1179648), 
      32138(147456), 6026(27648), 1130(5184), 212(972), 280(1296) ], 
  [ 73(162), 3520(7776), 440(972), 292(648), 1552(3456), 8272(18432), 
      1034(2304), 194(432) ], 
  [ 77(162), 3712(7776), 19792(41472), 2474(5184), 464(972), 616(1296), 
      3280(6912), 17488(36864), 2186(4608), 410(864) ], 
  [ 89(162), 4288(7776), 22864(41472), 121936(221184), 650320(1179648), 
      81290(147456), 15242(27648), 2858(5184), 536(972), 712(1296) ], 
  [ 127(162), 6112(7776), 764(972), 508(648), 2704(3456), 14416(18432), 
      1802(2304), 338(432) ], 
  [ 14(216), 22(324), 112(1728), 592(9216), 74(1152) ], 
  [ 86(216), 130(324), 688(1728), 3664(9216), 458(1152) ] ]
gap> List(cycsrc,Length);
[ 1, 3, 4, 6, 4, 1, 1, 6, 8, 8, 6, 8, 8, 3, 3, 10, 10, 8, 10, 10, 8, 5, 
  5 ]
gap> Sum(List(Flat(cycsrc),Density));
5097073/5308416
gap> Float(last); # already about 96% 'coverage'
0.960187
]]>
</Example>

There are also some elements of infinite order whose cycles seem to be
all finite, but <Q>on average</Q> pretty <Q>long</Q> -- e.g.

<Example>
<![CDATA[
gap> g := (b*a*c)^2*a;;
gap> Display(g);

Rcwa permutation of Z with modulus 288

        /
        | (16n-1)/3   if n in 1(3)
        | (9n+5)/4    if n in 3(24) U 11(24)
        | (27n+19)/4  if n in 15(24) U 23(24)
        | (3n+1)/4    if n in 5(24)
        | (n-3)/6     if n in 21(24)
        | (27n+29)/8  if n in 9(48) U 41(48)
        | (9n+7)/8    if n in 17(48) U 33(48)
        | (2n-7)/9    if n in 8(36)
 n |-> <  (4n-11)/9   if n in 32(36)
        | (27n+38)/8  if n in 14(48)
        | (3n+2)/8    if n in 26(48)
        | (9n+10)/8   if n in 38(48)
        | (3n+4)/4    if n in 20(72)
        | n/4         if n in 56(72)
        | (9n+14)/16  if n in 2(96)
        | (27n+58)/16 if n in 50(96)
        | n           if n in 0(6)
        \

gap> List([1..100],n->Length(Cycle(g,n)));
[ 6, 1, 6, 6, 6, 1, 194, 6, 216, 26, 26, 1, 26, 194, 65, 26, 26, 1, 216, 
  26, 6, 216, 46, 1, 640, 26, 70, 194, 216, 1, 70, 26, 216, 216, 26, 1, 
  194, 216, 73, 26, 110, 1, 194, 216, 194, 111, 39, 1, 194, 640, 640, 
  194, 26, 1, 171, 194, 204, 640, 216, 1, 111, 70, 91, 26, 194, 1, 216, 
  216, 26, 111, 65, 1, 50, 194, 26, 216, 640, 1, 502, 26, 111, 40, 110, 
  1, 26, 194, 385, 640, 88, 1, 100, 111, 65, 110, 416, 1, 171, 194, 194, 
  640 ]
gap> Length(Cycle(g,25));
640
gap> Maximum(Cycle(g,25));
323270249684063829
gap> Length(Cycle(g,25855));
4751
gap> Maximum(Cycle(g,25855));
507359605810239426786254778159924369135184044618585904603866210104085
gap> cycs := ShortCycles(g,[0..50000],10000,10^100);;
gap> S := [0..50000];;
gap> for cyc in cycs do S := Difference(S,cyc); od;
gap> S; # no cycle containing some n in [0..50000] has length > 10000 
[  ]
]]>
</Example>

Taking a look at the lengths of the trajectories of the Collatz mapping
<M>T</M> starting at the points in a cycle, we can see how a cycle of
<M>g</M> goes <Q>up and down</Q> in the <M>3n+1</M> tree:

<Example>
<![CDATA[
gap> List(Cycle(g,25),n->Length(Trajectory(T,n,[1])));
[ 17, 21, 25, 29, 33, 31, 35, 34, 32, 33, 37, 41, 45, 44, 42, 39, 43, 
  41, 45, 44, 42, 43, 40, 38, 35, 39, 37, 41, 40, 44, 48, 46, 50, 49, 
  47, 48, 45, 42, 46, 44, 48, 47, 45, 46, 50, 49, 47, 43, 41, 38, 39, 
  36, 34, 30, 27, 31, 29, 33, 32, 30, 31, 35, 33, 37, 36, 40, 39, 43, 
  41, 45, 44, 42, 43, 47, 51, 55, 53, 57, 56, 54, 55, 59, 58, 62, 66, 
  64, 68, 67, 65, 66, 63, 60, 64, 62, 66, 65, 63, 64, 68, 67, 65, 61, 
  59, 56, 52, 49, 53, 51, 55, 54, 52, 53, 57, 55, 59, 58, 56, 57, 54, 
  50, 48, 45, 49, 47, 51, 50, 54, 52, 56, 55, 53, 54, 58, 62, 66, 70, 
  74, 72, 76, 75, 79, 83, 87, 91, 90, 94, 93, 97, 95, 99, 98, 96, 97, 
  94, 91, 88, 85, 89, 87, 91, 90, 88, 89, 86, 84, 81, 85, 83, 87, 86, 
  90, 94, 98, 97, 101, 105, 109, 107, 111, 110, 108, 109, 113, 117, 115, 
  119, 118, 122, 126, 125, 123, 120, 124, 122, 126, 125, 123, 124, 121, 
  119, 116, 117, 114, 111, 115, 113, 117, 116, 114, 115, 119, 123, 122, 
  120, 117, 121, 119, 123, 122, 120, 121, 118, 116, 112, 110, 106, 103, 
  107, 105, 109, 108, 106, 107, 111, 109, 113, 112, 116, 114, 118, 117, 
  115, 116, 113, 110, 111, 108, 104, 102, 99, 103, 101, 105, 104, 108, 
  106, 110, 109, 107, 108, 112, 111, 109, 105, 102, 103, 100, 98, 95, 
  92, 96, 94, 98, 97, 95, 96, 93, 91, 88, 92, 90, 94, 93, 97, 101, 105, 
  109, 108, 106, 103, 107, 105, 109, 108, 106, 107, 104, 102, 99, 103, 
  101, 105, 104, 108, 112, 110, 114, 113, 111, 112, 116, 115, 113, 109, 
  106, 110, 108, 112, 111, 109, 110, 114, 112, 116, 115, 113, 114, 111, 
  107, 105, 102, 103, 100, 98, 95, 99, 97, 101, 100, 104, 103, 107, 105, 
  109, 108, 106, 107, 104, 101, 98, 99, 96, 94, 91, 92, 89, 87, 84, 85, 
  82, 80, 77, 81, 79, 83, 82, 86, 85, 89, 88, 86, 83, 80, 81, 78, 76, 
  73, 74, 71, 68, 72, 70, 74, 73, 71, 72, 76, 80, 79, 83, 87, 91, 90, 
  88, 85, 89, 87, 91, 90, 88, 89, 86, 84, 81, 85, 83, 87, 86, 90, 94, 
  92, 96, 95, 93, 94, 98, 96, 100, 99, 97, 98, 102, 106, 110, 114, 113, 
  111, 108, 112, 110, 114, 113, 111, 112, 109, 107, 104, 108, 106, 110, 
  109, 113, 117, 115, 119, 118, 116, 117, 114, 111, 115, 113, 117, 116, 
  114, 115, 119, 118, 116, 112, 110, 107, 108, 105, 103, 100, 104, 102, 
  106, 105, 109, 108, 112, 110, 114, 113, 111, 112, 116, 115, 113, 109, 
  106, 103, 104, 101, 99, 95, 91, 88, 92, 90, 94, 93, 91, 92, 96, 94, 
  98, 97, 95, 96, 100, 98, 102, 101, 105, 104, 102, 99, 100, 97, 93, 89, 
  87, 84, 85, 82, 80, 77, 74, 78, 76, 80, 79, 77, 78, 75, 73, 69, 67, 
  64, 68, 66, 70, 69, 73, 71, 75, 74, 72, 73, 70, 67, 68, 65, 63, 60, 
  64, 62, 66, 65, 69, 68, 66, 63, 64, 61, 59, 56, 60, 58, 62, 61, 65, 
  64, 62, 59, 60, 57, 55, 51, 48, 49, 46, 44, 40, 37, 34, 35, 32, 28, 
  26, 23, 27, 25, 29, 28, 32, 30, 34, 33, 31, 32, 36, 35, 33, 29, 26, 
  27, 24, 22, 19, 23, 21, 25, 24, 28, 27, 25, 22, 23, 20, 18, 14, 18, 
  22, 20, 24, 23, 21, 22, 19, 16, 20, 18, 22, 21, 19, 20, 24, 23, 21, 
  17, 15, 17, 15, 19, 18, 16 ]
gap> lngs := List(Cycle(g,25855),n->Length(Trajectory(T,n,[1])));;
gap> Minimum(lngs);
55
gap> Maximum(lngs);
521
gap> Position(lngs,55);
15
gap> Position(lngs,521);
2807
]]>
</Example>

Finally let's have a look at elements of <M>G</M> with small modulus:

<Example>
<![CDATA[
gap> B := RestrictedBall(G,One(G),20,36:Spheres);;
gap> List(B,Length);
[ 1, 3, 6, 12, 4, 6, 6, 4, 4, 4, 6, 6, 3, 3, 2, 0, 0, 0, 0, 0, 0 ]
gap> Sum(last);
70
gap> Position(last2,0)-2;
14
]]>
</Example>

So we have 70 elements of modulus 36 or less in <M>G</M> which
can be reached from the identity by successive multiplication with
generators without passing elements with mudulus exceeding 36.
Further we see that the longest word in the generators yielding an
element with modulus at most 36 has length&nbsp;14.
Now we double our bound on the modulus:

<Example>
<![CDATA[
gap> B := RestrictedBall(G,One(G),100,72:Spheres);;
gap> List(B,Length);
[ 1, 3, 6, 12, 22, 14, 18, 22, 24, 26, 26, 34, 35, 32, 37, 38, 46, 58, 
  65, 73, 82, 91, 93, 96, 110, 121, 114, 117, 146, 138, 148, 168, 174, 
  196, 215, 214, 232, 255, 280, 305, 315, 359, 377, 371, 363, 366, 397, 
  419, 401, 405, 405, 401, 407, 415, 435, 424, 401, 359, 338, 330, 332, 
  281, 278, 271, 269, 254, 255, 257, 258, 258, 233, 215, 202, 185, 154, 
  121, 88, 55, 35, 20, 10, 5, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
  0, 0, 0, 0, 0 ]
gap> Sum(last);
15614
gap> Position(last2,0)-2;
83
gap> Collected(List(Flat(B),Modulus));
[ [ 1, 1 ], [ 6, 3 ], [ 12, 4 ], [ 18, 2 ], [ 24, 4 ], [ 36, 56 ], 
  [ 48, 4 ], [ 72, 15540 ] ]
]]>
</Example>

We observe that there are 15540 elements in <M>G</M> with modulus 72
which are <Q>reachable</Q> from the identity by successive
multiplication with generators without passing elements with mudulus
exceeding 72. Further we see that the longest word in the generators
yielding an element with modulus at most 72 has length&nbsp;83. <P/>

It is obvious that many questions regarding the group <M>G</M>
remain open.

</Section>

<!-- #################################################################### -->

<Section Label="sec:GroupWithHugeFiniteOrbits">
<Heading>
  A group with huge finite orbits
</Heading>

In this section we investigate a group which has huge finite orbits on &ZZ;.

<Example>
<![CDATA[
gap> a := ClassTransposition(0,2,1,2);;
gap> b := ClassTransposition(0,5,4,5);;
gap> c := ClassTransposition(1,4,0,6);;
gap> G := Group(a,b,c);
<(0(2),1(2)),(0(5),4(5)),(1(4),0(6))>
gap> LoadDatabaseOfGroupsGeneratedBy3ClassTranspositions();
"3CTsGroups6"
gap> 3CTsGroups6.Id3CTsGroup(G,3CTsGroups6.grps); # 'catalogue number' of G
1284
]]>
</Example>

We look for orbits of length at most 100 containing an integer in the
range <C>[0..1000]</C>:

<Example>
<![CDATA[
gap> orbs := ShortOrbits(G,[0..1000],100);;
gap> List(orbs,Length);
[ 16, 2, 24, 2, 2, 2, 8, 2, 8, 2, 2, 8, 2, 8, 2, 2, 2, 40, 2, 8, 24, 2, 
  8, 2, 2, 8, 2, 24, 8, 2, 56, 2, 2, 2, 8, 2, 8, 2, 2, 8, 2, 8, 2, 2, 2, 
  24, 2, 8, 2, 8, 2, 2, 8, 2, 8, 2, 24, 2, 2, 2, 8, 2, 8, 2, 2, 8, 2, 8, 
  2, 2, 2, 2, 8, 24, 2, 8, 2, 2, 8, 2, 24, 8, 2, 2, 2, 2, 8, 2, 8, 2, 2, 
  8, 2, 8, 2, 2, 2, 24, 2, 8, 2, 8, 2, 2, 8, 2, 8, 2, 24, 2, 2 ]
gap> Collected(last);
[ [ 2, 67 ], [ 8, 32 ], [ 16, 1 ], [ 24, 9 ], [ 40, 1 ], [ 56, 1 ] ]
gap> Length(Difference([0..1000],Union(orbs)));
491
]]>
</Example>

So almost half of the integers in the range <C>[0..1000]</C> lie in
orbits of length larger than 100. In fact there are much larger orbits.
For example: 

<Example>
<![CDATA[
gap> B := Ball(G,32,500,OnPoints:Spheres);; # compute ball about 32
gap> Position(B,[]); # <> fail -> we have exhausted the orbit
354
gap> Sum(List(B,Length)); # the orbit length
6296
gap> Maximum(Flat(B)); # the largest integer in the orbit
3301636381609509797437679
gap> B := Ball(G,736,5000,OnPoints:Spheres);; # the same for 736 ...
gap> Position(B,[]);
2997
gap> Sum(List(B,Length)); # the orbit length for this time
495448
gap> Maximum(Flat(B));
2461374276522713949036151811903149785690151467356354652860276957152301465\
0546360696627187194849439881973442451686685024708652634593861146709752378\
847078493406287854573381920553713155967741550498839
]]>
</Example>

It seems that the cycles of <M>abc</M> completely traverse all orbits
of <M>G</M>, with the only exception of the orbit of 0. Let's check this
in the above examples:

<Example>
<![CDATA[
gap> g := a*b*c;;
gap> Display(g);

Rcwa permutation of Z with modulus 60

        /
        | n-1       if n in 3(30) U 9(30) U 17(30) U 23(30) U 27(30) U 
        |                   29(30)
        | 3n/2      if n in 0(20) U 12(20) U 16(20)
        | n+1       if n in 2(20) U 6(20) U 10(20)
        | (2n+1)/3  if n in 7(30) U 13(30) U 19(30)
        | n+3       if n in 1(30) U 11(30)
 n |-> <  n-5       if n in 15(30) U 25(30)
        | (3n+12)/2 if n in 4(20)
        | (3n-12)/2 if n in 8(20)
        | n+5       if n in 14(20)
        | n-3       if n in 18(20)
        | (2n-7)/3  if n in 5(30)
        | (2n+9)/3  if n in 21(30)
        \

gap> Length(Cycle(g,32));
6296
gap> Length(Cycle(g,736));
495448
]]>
</Example>

Representatives and lengths of the cycles of <M>g</M> which intersect
nontrivially with the range <C>[0..1000]</C> are as follows:

<Example>
<![CDATA[
gap> CycleRepresentativesAndLengths(g,[0..1000]:notify:=50000);
n = 736: after 50000 steps, the iterate has 157 binary digits.
n = 736: after 100000 steps, the iterate has 135 binary digits.
n = 736: after 150000 steps, the iterate has 131 binary digits.
n = 736: after 200000 steps, the iterate has 507 binary digits.
n = 736: after 250000 steps, the iterate has 414 binary digits.
n = 736: after 300000 steps, the iterate has 457 binary digits.
n = 736: after 350000 steps, the iterate has 465 binary digits.
n = 736: after 400000 steps, the iterate has 325 binary digits.
n = 736: after 450000 steps, the iterate has 534 binary digits.
n = 896: after 50000 steps, the iterate has 359 binary digits.
n = 896: after 100000 steps, the iterate has 206 binary digits.
[ [ 1, 15 ], [ 2, 2 ], [ 16, 24 ], [ 22, 2 ], [ 26, 2 ], [ 32, 6296 ], 
  [ 46, 2 ], [ 52, 8 ], [ 56, 296 ], [ 62, 2 ], [ 76, 8 ], [ 82, 2 ], 
  [ 86, 2 ], [ 92, 8 ], [ 106, 2 ], [ 112, 104 ], [ 116, 8 ], 
  [ 122, 2 ], [ 136, 440 ], [ 142, 2 ], [ 146, 2 ], [ 152, 40 ], 
  [ 166, 2 ], [ 172, 8 ], [ 176, 24 ], [ 182, 2 ], [ 196, 8 ], 
  [ 202, 2 ], [ 206, 2 ], [ 212, 8 ], [ 226, 2 ], [ 232, 24 ], 
  [ 236, 8 ], [ 242, 2 ], [ 256, 56 ], [ 262, 2 ], [ 266, 2 ], 
  [ 272, 408 ], [ 286, 2 ], [ 292, 8 ], [ 296, 104 ], [ 302, 2 ], 
  [ 316, 8 ], [ 322, 2 ], [ 326, 2 ], [ 332, 8 ], [ 346, 2 ], 
  [ 352, 264 ], [ 356, 8 ], [ 362, 2 ], [ 376, 1304 ], [ 382, 2 ], 
  [ 386, 2 ], [ 392, 24 ], [ 406, 2 ], [ 412, 8 ], [ 416, 200 ], 
  [ 422, 2 ], [ 436, 8 ], [ 442, 2 ], [ 446, 2 ], [ 452, 8 ], 
  [ 466, 2 ], [ 472, 104 ], [ 476, 8 ], [ 482, 2 ], [ 496, 24 ], 
  [ 502, 2 ], [ 506, 2 ], [ 512, 696 ], [ 526, 2 ], [ 532, 8 ], 
  [ 536, 3912 ], [ 542, 2 ], [ 556, 8 ], [ 562, 2 ], [ 566, 2 ], 
  [ 572, 8 ], [ 586, 2 ], [ 592, 888 ], [ 596, 8 ], [ 602, 2 ], 
  [ 616, 728 ], [ 622, 2 ], [ 626, 2 ], [ 632, 2776 ], [ 646, 2 ], 
  [ 652, 8 ], [ 656, 24 ], [ 662, 2 ], [ 676, 8 ], [ 682, 2 ], 
  [ 686, 2 ], [ 692, 8 ], [ 706, 2 ], [ 712, 24 ], [ 716, 8 ], 
  [ 722, 2 ], [ 736, 495448 ], [ 742, 2 ], [ 746, 2 ], [ 752, 1272 ], 
  [ 766, 2 ], [ 772, 8 ], [ 776, 376 ], [ 782, 2 ], [ 796, 8 ], 
  [ 802, 2 ], [ 806, 2 ], [ 812, 8 ], [ 826, 2 ], [ 832, 120 ], 
  [ 836, 8 ], [ 842, 2 ], [ 856, 2264 ], [ 862, 2 ], [ 866, 2 ], 
  [ 872, 24 ], [ 886, 2 ], [ 892, 8 ], [ 896, 132760 ], [ 902, 2 ], 
  [ 916, 8 ], [ 922, 2 ], [ 926, 2 ], [ 932, 8 ], [ 946, 2 ], 
  [ 952, 456 ], [ 956, 8 ], [ 962, 2 ], [ 976, 24 ], [ 982, 2 ], 
  [ 986, 2 ], [ 992, 1064 ] ]
]]>
</Example>

So far the author has checked that all positive integers less than
173176 lie in finite cycles of <M>g</M>. Several of them are longer
than 1000000, and the cycle containing 25952 has length 245719352.
Whether the cycle containing 173176 is finite or infinite has not
been checked so far -- in any case it is longer than 5700000000, and
it exceeds <M>10^{40000}</M>. Presumably it is finite as well, but
checking this may require a lot of computing time. <P/>

On the one hand the cycles of <M>g</M> seem to behave <Q>randomly</Q>,
perhaps as if they would ascend or descend from one point to the next
by a certain factor depending on which side a thrown coin falls on.
-- In this <Q>model</Q>, cycles would be finite with probability 1
since the simple random walk on &ZZ; is recurrent.
On the other, there seems to be quite some structure on them,
however little is known so far. <P/>

First we observe that each orbit under the action of <M>G</M>
seems to split into two cycles of <M>h := abcacb</M> of the same
length (of course this has been checked for many more orbits than
those shown here):

<Example>
<![CDATA[
gap> h := a*b*c*a*c*b;
<rcwa permutation of Z with modulus 360>
gap> List(CyclesOnFiniteOrbit(G,h,32),Length);
[ 3148, 3148 ]
gap> List(CyclesOnFiniteOrbit(G,h,736),Length);
[ 247724, 247724 ]
]]>
</Example>

One cycle seems to contain the points at the odd positions and the
other seems to contain the points at the even positions in the
cycle of <M>g</M>:

<Example>
<![CDATA[
gap> cycle_g := Cycle(g,32);;
gap> positions1 := List(Cycle(h,32),n->Position(cycle_g,n));;
gap> Collected(positions1 mod 2);
[ [ 1, 3148 ] ]
gap> positions2 := List(Cycle(h,33),n->Position(cycle_g,n));;
gap> Collected(positions2 mod 2);
[ [ 0, 3148 ] ]
]]>
</Example>

However the ordering in which these points are traversed looks
pretty <Q>scrambled</Q>:

<Example>
<![CDATA[
gap> positions1{[1..200]};
[ 1, 6271, 6291, 6281, 6285, 6287, 6283, 6289, 6273, 6275, 6277, 6279, 
  6293, 5, 15, 17, 19, 6259, 6261, 6263, 6265, 21, 23, 25, 41, 6227, 
  6229, 6231, 6233, 6235, 6237, 6239, 43, 53, 55, 57, 63, 59, 61, 65, 
  45, 47, 49, 51, 67, 6223, 6221, 69, 6163, 6215, 6205, 6209, 6211, 
  6207, 6213, 6165, 6171, 6177, 6179, 6181, 6183, 6175, 6173, 6185, 
  6189, 6191, 6187, 6193, 6169, 6167, 6195, 6199, 6201, 6197, 6203, 
  6217, 73, 83, 85, 87, 103, 113, 115, 117, 4357, 4361, 4363, 4359, 
  4365, 4371, 4373, 4375, 4377, 4369, 4367, 4379, 119, 121, 123, 125, 
  129, 131, 127, 133, 139, 141, 143, 145, 137, 135, 147, 149, 151, 153, 
  155, 159, 161, 157, 163, 169, 175, 4283, 4281, 177, 4271, 4273, 4275, 
  4277, 181, 4255, 4257, 4259, 4261, 4263, 4265, 4267, 183, 2161, 2163, 
  4195, 4199, 4201, 4197, 4203, 4209, 4211, 4213, 4215, 4207, 4205, 
  4217, 2165, 2167, 2169, 2171, 2175, 2177, 2173, 2179, 2185, 2187, 
  2189, 2191, 2183, 2181, 2193, 2195, 2197, 2199, 2201, 2467, 2469, 
  4117, 4121, 4123, 4119, 4125, 4131, 4133, 4135, 4137, 4129, 4127, 
  4139, 2471, 2473, 2475, 2477, 2487, 2489, 2491, 2507, 2517, 2519, 
  2521, 2537, 3923, 3925, 3941, 3943 ]
]]>
</Example>

</Section>

<!-- #################################################################### -->

<Section Label="sec:FourTransitiveGroup">
<Heading>
  A group which acts 4-transitively on the positive integers
</Heading>

In this section, we would like to show that the group <M>G</M> generated
by the two permutations

<Example>
<![CDATA[
gap> a := RcwaMapping([[3,0,2],[3,1,4],[3,0,2],[3,-1,4]]);;
gap> u := RcwaMapping([[3,0,5],[9,1,5],[3,-1,5],[9,-2,5],[9,4,5]]);;
gap> SetName(a,"a"); SetName(u,"u"); G := Group(a,u);;
]]>
</Example>

which we have already investigated in earlier examples acts 4-transitively
on the set of positive integers.

Obviously, it acts on the set of positive integers.
First we show that this action is transitive.

We start by checking in which residue classes sufficiently large positive
integers are mapped to smaller ones by a suitable group element:

<Example>
<![CDATA[
gap> List([a,a^-1,u,u^-1],DecreasingOn);
[ 1(2), 0(3), 0(5) U 2(5), 2(3) ]
gap> Union(last);
Z \ 4(30) U 16(30) U 28(30)
]]>
</Example>

We see that we cannot always choose such a group element from the set of
generators and their inverses -- otherwise the union would be
<C>Integers</C>.

<Example>
<![CDATA[
gap> List([a,a^-1,u,u^-1,a^2,a^-2,u^2,u^-2],DecreasingOn);
[ 1(2), 0(3), 0(5) U 2(5), 2(3), 1(8) U 7(8), 0(3) U 2(9) U 7(9), 
  0(25) U 12(25) U 17(25) U 20(25), 2(3) U 1(9) U 3(9) ]
gap> Union(last); # Still not enough ...
Z \ 4(90) U 58(90) U 76(90)
gap> List([a,a^-1,u,u^-1,a^2,a^-2,u^2,u^-2,a*u,u*a,(a*u)^-1,(u*a)^-1],
>         DecreasingOn);
[ 1(2), 0(3), 0(5) U 2(5), 2(3), 1(8) U 7(8), 0(3) U 2(9) U 7(9), 
  0(25) U 12(25) U 17(25) U 20(25), 2(3) U 1(9) U 3(9), 
  3(5) U 0(10) U 7(20) U 9(20), 0(5) U 2(5), 2(3), 3(9) U 4(9) U 8(9) ]
gap> Union(last); # ... but that's it!
Integers
]]>
</Example>

Finally, we have to deal with <Q>small</Q> integers. We use the notation for
the coefficients of rcwa mappings introduced at the beginning of this manual.
Let <M>c_{r(m)} &gt; a_{r(m)}</M>. Then we easily see that
<M>(a_{r(m)}n+b_{r(m)})/c_{r(m)} &gt; n</M> implies
<M>n &lt; b_{r(m)}/(c_{r(m)}-a_{r(m)})</M>.
Thus we can restrict our considerations to integers
<M>n &lt; b_{\rm max}</M>, where <M>b_{\rm max}</M> is the largest
second entry of a coefficient triple of one of the group elements
in our list:

<Example>
<![CDATA[
gap> List([a,a^-1,u,u^-1,a^2,a^-2,u^2,u^-2,a*u,u*a,(a*u)^-1,(u*a)^-1],
>         f->Maximum(List(Coefficients(f),c->c[2])));
[ 1, 1, 4, 2, 7, 7, 56, 28, 25, 17, 17, 11 ]
gap> Maximum(last);
56
]]>
</Example>

Thus this upper bound is 56. The rest is easy -- all we have to do is
to check that the orbit containing&nbsp;1 contains also all other positive
integers less than or equal to&nbsp;56:

<Example>
<![CDATA[
gap> S := [1];;
gap> while not IsSubset(S,[1..56]) do
>      S := Union(S,S^a,S^u,S^(a^-1),S^(u^-1));
>    od;
gap> IsSubset(S,[1..56]);
true
]]>
</Example>

Checking 2-transitivity is computationally harder, and in the sequel we
will omit some steps which are in practice needed to find out
<Q>what&nbsp;to&nbsp;do</Q>.

The approach taken here is to show that the stabilizer of&nbsp;1
in&nbsp;<M>G</M> acts transitively on the set of positive integers greater
than&nbsp;1. We do this by similar means as used above for showing the
transitivity of the action of <M>G</M> on the positive integers.

We start by determining all products of at most 5 generators and their
inverses, which stabilize&nbsp;1 (taking at most 4-generator products would
not suffice!):

<Example>
<![CDATA[
gap> gens := [a,u,a^-1,u^-1];;
gap> tups := Concatenation(List([1..5],k->Tuples([1..4],k)));;
gap> Length(tups);
1364
gap> tups := Filtered(tups,tup->ForAll([[1,3],[3,1],[2,4],[4,2]],
>                                      l->PositionSublist(tup,l)=fail));;
gap> Length(tups);
484
gap> stab := [];;
gap> for tup in tups do
>      n := 1;
>      for i in tup do n := n^gens[i]; od;
>      if n = 1 then Add(stab,tup); fi;
>    od;
gap> Length(stab);
118
gap> stabelm := List(stab,tup->Product(List(tup,i->gens[i])));;
gap> ForAll(stabelm,elm->1^elm=1); # Check.
true
]]>
</Example>

The resulting products have various different not quite small moduli:

<Example>
<![CDATA[
gap> List(stabelm,Modulus);
[ 4, 3, 16, 25, 9, 81, 64, 100, 108, 100, 25, 75, 27, 243, 324, 243, 
  256, 400, 144, 400, 100, 432, 324, 400, 80, 400, 625, 25, 75, 135, 
  150, 75, 225, 81, 729, 486, 729, 144, 144, 81, 729, 1296, 729, 6561, 
  1024, 1600, 192, 1600, 400, 576, 432, 1600, 320, 1600, 2500, 100, 100, 
  180, 192, 192, 108, 972, 1728, 972, 8748, 1600, 400, 320, 80, 1600, 
  2500, 300, 2500, 625, 625, 75, 675, 75, 75, 135, 405, 600, 120, 600, 
  1875, 75, 225, 405, 225, 225, 675, 243, 2187, 729, 2187, 216, 216, 
  243, 2187, 1944, 2187, 19683, 576, 144, 576, 432, 81, 81, 729, 2187, 
  5184, 324, 8748, 243, 2187, 19683, 26244, 19683 ]
gap> Lcm(last);
12597120000
gap> Collected(Factors(last));
[ [ 2, 10 ], [ 3, 9 ], [ 5, 4 ] ]
]]>
</Example>

Similar as before, we determine for any of the above mappings the
residue classes whose elements larger than the largest <M>b_{r(m)}</M>
- coefficient of the respective mapping are mapped to smaller integers:

<Example>
<![CDATA[
gap> decs := List(stabelm,DecreasingOn);;
gap> List(decs,Modulus);
[ 2, 3, 8, 25, 9, 9, 16, 100, 12, 50, 25, 75, 27, 81, 54, 81, 64, 400, 
  48, 200, 100, 72, 108, 400, 80, 200, 625, 25, 75, 45, 75, 75, 225, 81, 
  243, 81, 243, 144, 144, 81, 243, 216, 243, 243, 128, 1600, 64, 400, 
  400, 48, 144, 1600, 320, 400, 2500, 100, 100, 60, 96, 192, 108, 324, 
  144, 324, 972, 400, 400, 80, 80, 400, 2500, 100, 1250, 625, 625, 25, 
  75, 75, 75, 45, 135, 600, 120, 150, 1875, 75, 225, 135, 225, 225, 675, 
  243, 729, 243, 729, 108, 216, 243, 729, 162, 729, 2187, 144, 144, 144, 
  144, 81, 81, 243, 729, 1296, 324, 972, 243, 729, 2187, 1458, 2187 ]
gap> Lcm(last);
174960000
]]>
</Example>

Since the least common multiple of the moduli of these unions of residue
classes is as large as 174960000, directly forming their union and
checking whether it is equal to the set of integers would take relatively
much time and memory. However, starting with the set of integers and
subtracting the above sets one-by-one in a suitably chosen order is cheap:

<Example>
<![CDATA[
gap> SortParallel(decs,stabelm,
>      function(S1,S2)
>        return First([1..100],k->Factorial(k) mod Modulus(S1)=0)
>             < First([1..100],k->Factorial(k) mod Modulus(S2)=0);
>      end);
gap> S := Integers;;
gap> for i in [1..Length(decs)] do
>      S_old := S; S := Difference(S,decs[i]);
>      if S <> S_old then ViewObj(S); Print("\n"); fi;
>      if S = [] then maxind := i; break; fi;
>    od;
0(2)
2(6) U 4(6)
<union of 8 residue classes (mod 30)>
<union of 19 residue classes (mod 90) (9 classes)>
<union of 114 residue classes (mod 720)>
<union of 99 residue classes (mod 720)>
<union of 57 residue classes (mod 720)>
<union of 54 residue classes (mod 720)>
<union of 41 residue classes (mod 720)>
<union of 35 residue classes (mod 720)>
<union of 8 residue classes (mod 720) (6 classes)>
4(720) U 94(720) U 148(720) U 238(720)
<union of 24 residue classes (mod 5760)>
<union of 72 residue classes (mod 51840)>
<union of 48 residue classes (mod 51840)>
<union of 192 residue classes (mod 259200)>
<union of 168 residue classes (mod 259200)>
<union of 120 residue classes (mod 259200)>
<union of 96 residue classes (mod 259200)>
<union of 72 residue classes (mod 259200)>
<union of 60 residue classes (mod 259200)>
<union of 48 residue classes (mod 259200)>
<union of 24 residue classes (mod 259200)>
<union of 12 residue classes (mod 259200) (6 classes)>
<union of 24 residue classes (mod 777600)>
<union of 12 residue classes (mod 777600) (6 classes)>
111604(194400) U 14404(777600) U 208804(777600)
[  ]
]]>
</Example>

Similar as above, it remains to check that the <Q>small</Q> integers all
lie in the orbit containing&nbsp;2. Obviously, it is sufficient to check
that any integer greater than&nbsp;2 is mapped to a smaller one by some
suitably chosen element of the stabilizer under consideration:

<Example>
<![CDATA[
gap> Maximum(List(stabelm{[1..maxind]},
>                 f->Maximum(List(Coefficients(f),c->c[2]))));
6581
gap> Filtered([3..6581],n->Minimum(List(stabelm,elm->n^elm))>=n);
[ 4 ]
]]>
</Example>

We have to treat 4 separately:

<Example>
<![CDATA[
gap> 1^(u*a*u^2*a^-1*u);
1
gap> 4^(u*a*u^2*a^-1*u);
3
]]>
</Example>

Now we know that any positive integer greater than&nbsp;1 lies in the same
orbit under the action of the stabilizer of&nbsp;1 in&nbsp;<M>G</M>
as&nbsp;2, thus that this stabilizer acts transitively on
<M>&NN; \setminus \{1\}</M>. But this means that we have established
the 2-transitivity of the action of <M>G</M> on&nbsp;&NN;. <P/>

In the following, we essentially repeat the above steps to show that
this action is indeed 3-transitive:

<Example>
<![CDATA[
gap> tups := Concatenation(List([1..6],k->Tuples([1..4],k)));;
gap> tups := Filtered(tups,tup->ForAll([[1,3],[3,1],[2,4],[4,2]],
>                                      l->PositionSublist(tup,l)=fail));;
gap> stab := [];;
gap> for tup in tups do
>      l := [1,2];
>      for i in tup do l := List(l,n->n^gens[i]); od;
>      if l = [1,2] then Add(stab,tup); fi;
>    od;
gap> Length(stab);
212
gap> stabelm := List(stab,tup->Product(List(tup,i->gens[i])));;
gap> decs := List(stabelm,DecreasingOn);;
gap> SortParallel(decs,stabelm,function(S1,S2)
>      return First([1..100],k->Factorial(k) mod Mod(S1)=0)
>           < First([1..100],k->Factorial(k) mod Mod(S2)=0); end);
gap> S := Integers;;
gap> for i in [1..Length(decs)] do
>      S_old := S; S := Difference(S,decs[i]);
>      if S <> S_old then ViewObj(S); Print("\n"); fi;
>      if S = [] then break; fi;
>    od;
Z \ 1(8) U 7(8)
<union of 151 residue classes (mod 240)>
<union of 208 residue classes (mod 720)>
<union of 51 residue classes (mod 720)>
<union of 45 residue classes (mod 720)>
<union of 39 residue classes (mod 720)>
<union of 33 residue classes (mod 720)>
<union of 23 residue classes (mod 720)>
<union of 19 residue classes (mod 720) (7 classes)>
<union of 17 residue classes (mod 720) (6 classes)>
<union of 16 residue classes (mod 720) (7 classes)>
<union of 14 residue classes (mod 720) (9 classes)>
<union of 8 residue classes (mod 720) (6 classes)>
<union of 7 residue classes (mod 720) (6 classes)>
238(360) U 4(720) U 148(720) U 454(720)
<union of 38 residue classes (mod 5760)>
<union of 37 residue classes (mod 5760)>
<union of 25 residue classes (mod 5760)>
<union of 21 residue classes (mod 5760)>
<union of 17 residue classes (mod 5760) (13 classes)>
<union of 16 residue classes (mod 5760) (12 classes)>
<union of 138 residue classes (mod 51840)>
<union of 48 residue classes (mod 51840)>
<union of 32 residue classes (mod 51840)>
<union of 20 residue classes (mod 51840) (14 classes)>
<union of 16 residue classes (mod 51840) (12 classes)>
<union of 68 residue classes (mod 259200)>
<union of 42 residue classes (mod 259200)>
<union of 32 residue classes (mod 259200)>
<union of 26 residue classes (mod 259200)>
<union of 25 residue classes (mod 259200)>
<union of 11 residue classes (mod 259200) (10 classes)>
<union of 10 residue classes (mod 259200) (9 classes)>
<union of 7 residue classes (mod 259200) (6 classes)>
13414(129600) U 2164(259200) U 66964(259200) U 228964(259200)
2164(259200) U 66964(259200) U 228964(259200)
[  ]
gap> Maximum(List(stabelm,f->Maximum(List(Coefficients(f),c->c[2]))));
515816
gap> smallnum := [4..515816];;
gap> for i in [1..Length(stabelm)] do
>      smallnum := Filtered(smallnum,n->n^stabelm[i]>=n);
>    od;
gap> smallnum;
[  ]
]]>
</Example>

The same for 4-transitivity:

<Example>
<![CDATA[
gap> tups := Concatenation(List([1..8],k->Tuples([1..4],k)));;
gap> tups := Filtered(tups,tup->ForAll([[1,3],[3,1],[2,4],[4,2]],
>                                      l->PositionSublist(tup,l)=fail));;
gap> stab := [];;
gap> for tup in tups do
>      l := [1,2,3];
>      for i in tup do l := List(l,n->n^gens[i]); od;
>      if l = [1,2,3] then Add(stab,tup); fi;
>    od;
gap> Length(stab);
528
gap> stabelm := [];;
gap> for i in [1..Length(stab)] do
>      elm := One(G);
>      for j in stab[i] do
>        if Modulus(elm) > 10000 then elm := fail; break; fi;
>        elm := elm * gens[j];
>      od;
>      if elm <> fail then Add(stabelm,elm); fi;
>    od;
gap> Length(stabelm);
334
gap> decs := List(stabelm,DecreasingOn);;
gap> SortParallel(decs,stabelm,
>      function(S1,S2)
>        return First([1..100],k->Factorial(k) mod Modulus(S1) = 0)
>             < First([1..100],k->Factorial(k) mod Modulus(S2) = 0);
>      end);
gap> S := Integers;;
gap> for i in [1..Length(decs)] do
>      S_old := S; S := Difference(S,decs[i]);
>      if S <> S_old then ViewObj(S); Print("\n"); fi;
>      if S = [] then maxind := i; break; fi;
>    od;
Z \ 1(8) U 7(8)
<union of 46 residue classes (mod 72)>
<union of 20 residue classes (mod 72) (8 classes)>
4(18)
<union of 28 residue classes (mod 576)>
<union of 22 residue classes (mod 576)>
<union of 21 residue classes (mod 576)>
40(72) U 4(144) U 94(144) U 346(576) U 418(576)
<union of 16 residue classes (mod 576) (6 classes)>
<union of 15 residue classes (mod 576) (6 classes)>
4(144) U 94(144) U 346(576) U 418(576)
<union of 30 residue classes (mod 5184)>
<union of 26 residue classes (mod 5184)>
<union of 6 residue classes (mod 1296)>
<union of 504 residue classes (mod 129600)>
<union of 324 residue classes (mod 129600)>
<union of 282 residue classes (mod 129600)>
<union of 239 residue classes (mod 129600)>
<union of 218 residue classes (mod 129600)>
<union of 194 residue classes (mod 129600)>
<union of 154 residue classes (mod 129600)>
<union of 97 residue classes (mod 129600)>
<union of 85 residue classes (mod 129600)>
<union of 77 residue classes (mod 129600)>
<union of 67 residue classes (mod 129600)>
<union of 125 residue classes (mod 259200)>
<union of 108 residue classes (mod 259200)>
<union of 107 residue classes (mod 259200)>
<union of 101 residue classes (mod 259200)>
<union of 100 residue classes (mod 259200)>
<union of 84 residue classes (mod 259200)>
<union of 80 residue classes (mod 259200)>
<union of 76 residue classes (mod 259200)>
<union of 70 residue classes (mod 259200)>
<union of 66 residue classes (mod 259200)>
<union of 54 residue classes (mod 259200)>
<union of 53 residue classes (mod 259200)>
<union of 47 residue classes (mod 259200)>
<union of 43 residue classes (mod 259200)>
<union of 31 residue classes (mod 259200)>
<union of 24 residue classes (mod 259200)>
<union of 23 residue classes (mod 259200)>
<union of 13 residue classes (mod 259200) (8 classes)>
57406(129600) U 115006(129600) U 192676(259200) U 250276(259200)
57406(129600) U 192676(259200) U 250276(259200) U 374206(388800)
57406(129600) U 192676(259200) U 250276(259200)
250276(259200) U 57406(388800) U 316606(388800) U 451876(777600)
316606(388800) U 451876(777600) U 509476(777600) U 768676(777600)
<union of 18 residue classes (mod 3110400) (6 classes)>
451876(777600) U 509476(777600) U 705406(777600) U 768676(777600)
 U 2649406(3110400)
451876(777600) U 705406(777600) U 768676(777600) U 2649406(3110400)
451876(777600) U 705406(777600) U 2649406(3110400)
705406(777600) U 2007076(3110400) U 2649406(3110400) U 2784676(3110400)
<union of 14 residue classes (mod 9331200) (8 classes)>
2260606(2332800) U 5759806(9331200) U 5895076(9331200) U 8227876(9331200)
4593406(6998400) U 15091006(27993600) U 17559076(27993600)
 U 24557476(27993600)
<union of 14 residue classes (mod 83980800) (8 classes)>
18590206(20995200) U 24557476(83980800) U 45552676(83980800)
 U 71078206(83980800)
[  ]
gap> Maximum(List(stabelm{[1..maxind]},
>                 f->Maximum(List(Coefficients(f),c->c[2]))));
58975
gap> smallnum := [5..58975];;
gap> for i in [1..maxind] do
>      smallnum := Filtered(smallnum,n->n^stabelm[i]>=n);
>    od;
gap> smallnum;
[  ]
]]>
</Example>

There is even some evidence that the degree of transitivity of the action of
<M>G</M> on the positive integers is higher than&nbsp;4:

<Example>
<![CDATA[
gap> phi := EpimorphismFromFreeGroup(G);
[ a, u ] -> [ a, u ]
gap> F := Source(phi);
<free group on the generators [ a, u ]>
gap> List([5..20],
>         n->RepresentativeActionPreImage(G,[1,2,3,4,5],
>                                           [1,2,3,4,n],OnTuples,F));
[ <identity ...>, a^-3*u^4*a*u^-2*a^2, a^-1*(a^-1*u)^4*a^-1*u^-1*a, 
  a^4*u^-2*a^-4, a^-1*u^-4*a, (u^2*a^-1)^2*u^-2, u^-2*a^-2*u^4, 
  a^-1*u^2*a, a^-1*u^-6*a, a^2*u^4*a^2*u^2, u^-4*a*u^-2*a^-3, 
  a^-1*u^-2*a^-3*u^4*a^2, a^2*(a*u^2)^2, (a*u^-4)^2*a^-2, 
  u^-2*a*u^2*a*u^-2, u^-4*a^2*u^2 ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzlikePerms);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:ThreeButNotFourTransitiveGroup">
<Heading>
  A group which acts 3-transitively, but not 4-transitively on &ZZ;
</Heading>

In this section, we would like to show that the group <M>G</M> generated
by the two permutations <M>n \mapsto n + 1</M> and <M>\tau_{1(2),0(4)}</M>
acts 3-transitively, but not 4-transitively on the set of integers.

<Example>
<![CDATA[
gap> G := Group(ClassShift(0,1),ClassTransposition(1,2,0,4));
<rcwa group over Z with 2 generators>
gap> IsTame(G);
false
gap> (G.1^-2*G.2)^3*(G.1^2*G.2)^3; # G <> the free product C_infty * C_2.
IdentityMapping( Integers )
gap> Display(G:CycleNotation:=false);

Wild rcwa group over Z, generated by

[
Tame rcwa permutation of Z: n -> n + 1

Rcwa permutation of Z with modulus 4, of order 2

        /
        | 2n-2    if n in 1(2)
 n |-> <  (n+2)/2 if n in 0(4)
        | n       if n in 2(4)
        \

]
]]>
</Example>

This group acts transitively on&nbsp;&ZZ;, since already the cyclic
group generated by the first of the two generators does so. Next we have to
show that it acts 2-transitively. We essentially proceed as in the example
in the previous section, by checking that the stabilizer of&nbsp;0
acts transitively on <M>&ZZ; \setminus \{0\}</M>.

<Example>
<![CDATA[
gap> gens := [ClassShift(0,1)^-1,ClassTransposition(1,2,0,4),
>             ClassShift(0,1)];;
gap> tups := Concatenation(List([1..6],k->Tuples([-1,0,1],k)));;
gap> tups := Filtered(tups,tup->ForAll([[0,0],[-1,1],[1,-1]],
>                                      l->PositionSublist(tup,l)=fail));;
gap> Length(tups);
189
gap> stab := [];;
gap> for tup in tups do
>      n := 0;
>      for i in tup do n := n^gens[i+2]; od;
>      if n = 0 then Add(stab,tup); fi;
>    od;
gap> stabelm := List(stab,tup->Product(List(tup,i->gens[i+2])));;
gap> Collected(List(stabelm,Modulus));
[ [ 4, 6 ], [ 8, 4 ], [ 16, 3 ] ]
gap> decs := List(stabelm,DecreasingOn);
[ 0(4), 3(4), 0(4), 3(4), 2(4), 0(4), 4(8), 2(4), 2(4), 0(4), 1(4), 
  0(8), 3(8) ]
gap> Union(decs);
Integers
]]>
</Example>

Similar as in the previous section, it remains to check that the integers
with <Q>small</Q> absolute value all lie in the orbit containing&nbsp;1 under
the action of the stabilizer of&nbsp;0:

<Example>
<![CDATA[
gap> Maximum(List(stabelm,f->Maximum(List(Coefficients(f),
>                                         c->AbsInt(c[2])))));
21
gap> S := [1];;
gap> for elm in stabelm do S := Union(S,S^elm,S^(elm^-1)); od;
gap> IsSubset(S,Difference([-21..21],[0])); # Not yet ..
false
gap> for elm in stabelm do S := Union(S,S^elm,S^(elm^-1)); od;
gap> IsSubset(S,Difference([-21..21],[0])); # ... but now!
true
]]>
</Example>

Now we have to check for 3-transitivity. Since we cannot find for every
residue class an element of the pointwise stabilizer of <M>\{0,1\}</M>
which properly divides its elements, we also have to take additions and
subtractions into consideration. Since the moduli of all of our stabilizer
elements are quite small, simply looking at sets of representatives is cheap:

<Example>
<![CDATA[
gap> tups := Concatenation(List([1..10],k->Tuples([-1,0,1],k)));;
gap> tups := Filtered(tups,tup->ForAll([[0,0],[-1,1],[1,-1]],
>                                      l->PositionSublist(tup,l)=fail));;
gap> Length(tups);
3069
gap> stab := [];;
gap> for tup in tups do
>      l := [0,1];
>      for i in tup do l := List(l,n->n^gens[i+2]); od;
>      if l = [0,1] then Add(stab,tup); fi;
>    od;
gap> Length(stab);
10
gap> stabelm := List(stab,tup->Product(List(tup,i->gens[i+2])));;
gap> Maximum(List(stabelm,Modulus));
8
gap> Maximum(List(stabelm,
>                 f->Maximum(List(Coefficients(f),c->AbsInt(c[2])))));
8
gap> decsp := List(stabelm,elm->Filtered([9..16],n->n^elm<n));
[ [ 9, 13 ], [ 10, 12, 14, 16 ], [ 12, 16 ], [ 9, 13 ], [ 12, 16 ], 
  [ 9, 11, 13, 15 ], [ 9, 11, 13, 15 ], [ 12, 16 ], [ 12, 16 ], 
  [ 9, 11, 13, 15 ] ]
gap> Union(decsp);
[ 9, 10, 11, 12, 13, 14, 15, 16 ]
gap> decsm := List(stabelm,elm->Filtered([-16..-9],n->n^elm>n));
[ [ -15, -13, -11, -9 ], [ -16, -12 ], [ -16, -12 ], [ -15, -11 ], 
  [ -16, -14, -12, -10 ], [ -15, -11 ], [ -15, -11 ], 
  [ -16, -14, -12, -10 ], [ -16, -14, -12, -10 ], [ -15, -11 ] ]
gap> Union(decsm);
[ -16, -15, -14, -13, -12, -11, -10, -9 ]
gap> S := [2];;
gap> for elm in stabelm do S := Union(S,S^elm,S^(elm^-1)); od;
gap> IsSubset(S,Difference([-8..8],[0,1]));
true
]]>
</Example>

At this point we have established 3-transitivity.
It remains to check that the group <M>G</M> does not act 4-transitively.
We do this by checking that it is not transitive on 4-tuples (mod&nbsp;4).
Since <M>n</M>&nbsp;mod&nbsp;8 determines the image of <M>n</M> under
a generator of <M>G</M> (mod&nbsp;4), it suffices to compute (mod&nbsp;8):

<Example>
<![CDATA[
gap> orb := [[0,1,2,3]];;
gap> extend := function ()
>      local gen;
>      for gen in gens do
>        orb := Union(orb,List(orb,l->List(l,n->n^gen) mod 8));
>      od;
>    end;;
gap> repeat
>      old := ShallowCopy(orb);
>      extend(); Print(Length(orb),"\n");
>    until orb = old;
7
27
97
279
573
916
1185
1313
1341
1344
1344
gap> Length(Set(List(orb,l->l mod 4)));
120
gap> last < 4^4;
true
]]>
</Example>

This shows that <M>G</M> acts not 4-transitively on&nbsp;&ZZ;.
The corresponding calculation for 3-tuples looks as follows:

<Example>
<![CDATA[
gap> orb := [[0,1,2]];;
gap> repeat
>      old := ShallowCopy(orb);
>      extend(); Print(Length(orb),"\n");
>    until orb = old;
7
27
84
207
363
459
503
512
512
gap> Length(Set(List(orb,l->l mod 4)));
64
gap> last = 4^3;
true
]]>
</Example>

Needless to say that the latter kind of argumentation is not suitable
for proving, but only for disproving <M>k</M>-transitivity. 

</Section>

<!-- #################################################################### -->

<Section Label="sec:SlowlyContractingMapping">
<Heading>
  An rcwa mapping which seems to be contracting, but very slow
</Heading>

The iterates of an integer under the Collatz mapping <M>T</M> seem to
approach its contraction centre -- this is the finite set where all
trajectories end up after a finite number of steps -- rather quickly and
do not get very large before doing so (of course this is a purely heuristic
statement as the <M>3n+1</M> conjecture has not been proved so far!):

<Example>
<![CDATA[
gap> T := RcwaMapping([[1,0,2],[3,1,2]]);;
gap> S0 := LikelyContractionCentre(T,100,1000);
#I  Warning: `LikelyContractionCentre' is highly probabilistic.
The returned result can only be regarded as a rough guess.
See ?LikelyContractionCentre for more information.
[ -136, -91, -82, -68, -61, -55, -41, -37, -34, -25, -17, -10, -7, -5, 
  -1, 0, 1, 2 ]
gap> S0^T = S0; # This holds by definition of the contraction centre.
true
gap> List([1..30],n->Length(Trajectory(T,n,S0)));
[ 1, 1, 5, 2, 4, 6, 11, 3, 13, 5, 10, 7, 7, 12, 12, 4, 9, 14, 14, 6, 6, 
  11, 11, 8, 16, 8, 70, 13, 13, 13 ]
gap> Maximum(List([1..1000],n->Length(Trajectory(T,n,S0))));
113
gap> Maximum(List([1..1000],n->Maximum(Trajectory(T,n,S0))));
125252
]]>
</Example>

The following mapping seems to be contracting as well, but its trajectories
are much longer:

<Log>
<![CDATA[
gap> f6 := RcwaMapping([[ 1,0,6],[ 5, 1,6],[ 7,-2,6],
>                       [11,3,6],[11,-2,6],[11,-1,6]]);;
gap> Display(f6);

Rcwa mapping of Z with modulus 6

        /
        | n/6       if n in 0(6)
        | (5n+1)/6  if n in 1(6)
        | (7n-2)/6  if n in 2(6)
 n |-> <  (11n+3)/6 if n in 3(6)
        | (11n-2)/6 if n in 4(6)
        | (11n-1)/6 if n in 5(6)
        |
        \

gap> S0 := LikelyContractionCentre(f6,1000,100000);;
#I  Warning: `LikelyContractionCentre' is highly probabilistic.
The returned result can only be regarded as a rough guess.
See ?LikelyContractionCentre for more information.
gap> Trajectory(f6,25,S0);
[ 25, 21, 39, 72, 12, 2 ]
gap> List([1..100],n->Length(Trajectory(f6,n,S0)));
[ 1, 1, 3, 4, 1, 2, 3, 2, 1, 5, 7, 2, 8, 17, 3, 16, 1, 4, 17, 6, 5, 2, 
  5, 5, 6, 1, 4, 2, 15, 1, 1, 3, 2, 5, 13, 3, 2, 3, 4, 1, 8, 4, 4, 2, 7, 
  19, 23517, 3, 9, 3, 1, 18, 14, 2, 20, 23512, 14, 2, 6, 6, 1, 4, 19, 
  12, 23511, 8, 23513, 10, 1, 13, 13, 3, 1, 23517, 7, 20, 7, 9, 9, 6, 
  12, 8, 6, 18, 14, 23516, 31, 12, 23545, 4, 21, 19, 5, 1, 17, 17, 13, 
  19, 6, 23515 ]
gap> Maximum(Trajectory(f6,47,S0));
7363391777762473304431877054771075818733690108051469808715809256737742295\
45698886054
]]>
</Log>

Computing the trajectory of 3224 takes quite a while -- this trajectory
ascends to about <M>3 \cdot 10^{2197}</M>, before it approaches the fixed
point&nbsp;2 after 19949562 steps. <P/>

When constructing the mapping <C>f6</C>, the denominators of the
partial mappings have been chosen to be equal and the numerators have
been chosen to be numbers coprime to the common denominator, whose product
is just a little bit smaller than the <C>Modulus(f6)</C>th power of the
denominator. In the example we have <M>5 \cdot 7 \cdot 11^3 = 46585</M>
and <M>6^6 = 46656</M>. <P/>

Although the trajectories of <C>T</C> are much shorter than those of
<C>f6</C>, it seems likely that this does not make the problem of deciding
whether the mapping&nbsp;<C>T</C> is contracting essentially easier --
even for mappings with much shorter trajectories than&nbsp;<C>T</C>
the problem seems to be equally hard. A solution can usually only be found
in trivial cases, i.e. for example when there is some <M>k</M> such that
applying the <M>k</M>th power of the respective mapping to any integer
decreases its absolute value. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().SlowlyContractingMappings);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:AndaloroResult">
<Heading>Checking a result by P. Andaloro</Heading>

In <Cite Key="Andaloro00"/>, P.&nbsp;Andaloro has shown that proving that
trajectories of integers <M>n \in 1(16)</M> under the Collatz mapping always
contain&nbsp;1 would be sufficient to prove the <M>3n+1</M> conjecture.
In the sequel, this result is verified by &RCWA;. Checking that the
union of the images of the residue class 1(16) under powers of the Collatz
mapping&nbsp;<M>T</M> contains <M>&ZZ; \setminus 0(3)</M> is obviously
enough. Thus we put <M>S := 1(16)</M>, and successively unite the
set&nbsp;<M>S</M> with its image under&nbsp;<M>T</M>:

<Example>
<![CDATA[
gap> T := RcwaMapping([[1,0,2],[3,1,2]]);
<rcwa mapping of Z with modulus 2>
gap> S := ResidueClass(Integers,16,1);
1(16)
gap> S := Union(S,S^T);
1(16) U 2(24)
gap> S := Union(S,S^T);
1(12) U 2(24) U 17(48) U 33(48)
gap> S := Union(S,S^T);
<union of 30 residue classes (mod 144)>
gap> S := Union(S,S^T);
<union of 42 residue classes (mod 144)>
gap> S := Union(S,S^T);
<union of 172 residue classes (mod 432)>
gap> S := Union(S,S^T);
<union of 676 residue classes (mod 1296)>
gap> S := Union(S,S^T);
<union of 810 residue classes (mod 1296)>
gap> S := Union(S,S^T);
<union of 2638 residue classes (mod 3888)>
gap> S := Union(S,S^T);
<union of 33 residue classes (mod 48)>
gap> S := Union(S,S^T);
<union of 33 residue classes (mod 48)>
gap> Union(S,ResidueClass(Integers,3,0)); # Et voila ...
Integers
]]>
</Example>

Further similar computations are shown in
Section&nbsp;<Ref Label="sec:CollatzImagesAndPreImages"/>. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzMapping);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:MatthewsLeighExamples">
<Heading>Two examples by Matthews and Leigh</Heading>

In <Cite Key="MatthewsLeigh87"/>, K.&nbsp;R.&nbsp;Matthews and
G.&nbsp;M.&nbsp;Leigh have shown that two trajectories of the following
(surjective, but not injective) mappings are acyclic (mod&nbsp;<M>x</M>)
and divergent:

<Example>
<![CDATA[
gap> x := Indeterminate(GF(4),1);; SetName(x,"x");
gap> R := PolynomialRing(GF(2),1);
GF(2)[x]
gap> ML1 := RcwaMapping(R,x,[[1,0,x],[(x+1)^3,1,x]]*One(R));;
gap> ML2 := RcwaMapping(R,x,[[1,0,x],[(x+1)^2,1,x]]*One(R));;
gap> Display(ML1);

Rcwa mapping of GF(2)[x] with modulus x

        /
        | P/x                     if P in 0(x)
 P |-> <  ((x^3+x^2+x+1)*P + 1)/x if P in 1(x)
        |
        \

gap> Display(ML2);

Rcwa mapping of GF(2)[x] with modulus x

        /
        | P/x               if P in 0(x)
 P |-> <  ((x^2+1)*P + 1)/x if P in 1(x)
        |
        \

gap> List([ML1,ML2],IsSurjective);
[ true, true ]
gap> List([ML1,ML2],IsInjective);
[ false, false ]
gap> traj1 := Trajectory(ML1,One(R),16);
[ 1, x^2+x+1, x^4+x^2+x, x^3+x+1, x^5+x^4+x^2, x^4+x^3+x, x^3+x^2+1, 
  x^5+x^2+1, x^7+x^6+x^5+x^3+1, x^9+x^7+x^6+x^5+x^3+x+1, 
  x^11+x^10+x^8+x^7+x^6+x^5+x^2, x^10+x^9+x^7+x^6+x^5+x^4+x, 
  x^9+x^8+x^6+x^5+x^4+x^3+1, x^11+x^8+x^7+x^6+x^4+x+1, 
  x^13+x^12+x^11+x^8+x^7+x^6+x^4, x^12+x^11+x^10+x^7+x^6+x^5+x^3 ]
gap> traj2 := Trajectory(ML2,(x^3+x+1)*One(R),16);
[ x^3+x+1, x^4+x+1, x^5+x^3+x^2+x+1, x^6+x^3+1, x^7+x^5+x^4+x^2+x, 
  x^6+x^4+x^3+x+1, x^7+x^4+x^3+x+1, x^8+x^6+x^5+x^4+x^3+x+1, 
  x^9+x^6+x^3+x+1, x^10+x^8+x^7+x^5+x^4+x+1, 
  x^11+x^8+x^7+x^5+x^4+x^3+x^2+x+1, x^12+x^10+x^9+x^8+x^7+x^5+1, 
  x^13+x^10+x^7+x^4+x, x^12+x^9+x^6+x^3+1, 
  x^13+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x, 
  x^12+x^10+x^9+x^7+x^6+x^4+x^3+x+1 ]
]]>
</Example>

The pattern which Matthews and Leigh used to show the divergence of the
above trajectories can be recognized easily by looking at the corresponding
Markov chains with the two states 0&nbsp;mod&nbsp;<M>x</M> and
1&nbsp;mod&nbsp;<M>x</M>:

<Example>
<![CDATA[
gap> traj1modx := Trajectory(ML1,One(R),400,x);;
gap> traj2modx := Trajectory(ML2,(x^3+x+1)*One(R),600,x);;
gap> List(traj1modx{[1..150]},val->Position([Zero(R),One(R)],val)-1);
[ 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 
  1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 
  1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ]
gap> List(traj2modx{[1..150]},val->Position([Zero(R),One(R)],val)-1);
[ 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 
  1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 
  0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1 ]
]]>
</Example>

What is important here are the lengths of the intervals between two changes
from one state to the other:

<Example>
<![CDATA[
gap> ChangePoints := l->Filtered([1..Length(l)-1],pos->l[pos]<>l[pos+1]);;
gap> Diffs := l->List([1..Length(l)-1],pos->l[pos+1]-l[pos]);;
gap> Diffs(ChangePoints(traj1modx)); # The pattern in the first ...
[ 1, 1, 2, 4, 2, 2, 4, 8, 4, 4, 8, 16, 8, 8, 16, 32, 16, 16, 32, 64, 32, 
  32, 64 ]
gap> Diffs(ChangePoints(traj2modx)); # ... and in the second example.
[ 1, 7, 1, 1, 1, 13, 1, 1, 1, 1, 1, 1, 1, 25, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 49, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 97, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 193, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ]
gap> Diffs(ChangePoints(last)); # Make this a bit more obvious.
[ 1, 3, 1, 7, 1, 15, 1, 31, 1, 63, 1 ]
]]>
</Example>

This looks clearly acyclic, thus the trajectories diverge.
Needless to say however that this computational evidence does not replace
the proof along these lines given in the article cited above, but just
sheds a light on the idea behind it. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().MatthewsLeigh);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:OrdersOfCommutators">
<Heading>Orders of commutators</Heading>

We enter some wild rcwa permutation:

<Example>
<![CDATA[
gap> u := RcwaMapping([[3,0,5],[9,1,5],[3,-1,5],[9,-2,5],[9,4,5]]);;
gap> IsTame(u);;
gap> Display(u);

Wild rcwa permutation of Z with modulus 5

        /
        | 3n/5     if n in 0(5)
        | (9n+1)/5 if n in 1(5)
 n |-> <  (3n-1)/5 if n in 2(5)
        | (9n-2)/5 if n in 3(5)
        | (9n+4)/5 if n in 4(5)
        \
]]>
</Example>

We would like to compute the order of <M>[u,n \mapsto n + k]</M>
and <M>[u^2,n \mapsto n + k]</M> for different values of&nbsp;<M>k</M>:

<Example>
<![CDATA[
gap> nu := ClassShift(0,1);; # n -> n + 1
gap> l := Filtered([0..100],k->IsTame(Comm(u,nu^k)));
[ 0, 2, 3, 5, 6, 9, 10, 12, 13, 15, 17, 18, 20, 21, 24, 25, 27, 28, 30, 
  32, 33, 35, 36, 39, 40, 42, 43, 45, 47, 48, 50, 51, 54, 55, 57, 58, 
  60, 62, 63, 65, 66, 69, 70, 72, 73, 75, 77, 78, 80, 81, 84, 85, 87, 
  88, 90, 92, 93, 95, 96, 99, 100 ]
gap> List(l,k->Order(Comm(u,nu^k)));
[ 1, 6, 5, 3, 5, 5, 3, infinity, 7, infinity, 7, 5, 3, infinity, 
  infinity, 3, 5, 7, infinity, 7, infinity, 3, 5, 5, 3, 5, infinity, 
  infinity, infinity, 5, 3, 5, 5, 3, infinity, 7, infinity, 7, 5, 3, 
  infinity, infinity, 3, 5, 7, infinity, 7, infinity, 3, 5, 5, 3, 5, 
  infinity, infinity, infinity, 5, 3, 5, 5, 3 ]
gap> u2 := u^2;
<wild rcwa permutation of Z with modulus 25>
gap> Filtered([1..16],k->IsTame(Comm(u2,nu^k))); # k<15->[u^2,nu^k] wild!
[ 15 ]
gap> Order(Comm(u2,nu^15));
infinity
gap> u2nu17 := Comm(u2,nu^17);
<rcwa permutation of Z with modulus 81>
gap> cycs := ShortCycles(u2nu17,[-100..100],100);;
gap> List(cycs,Length);
[ 72, 73, 72, 72, 72, 73, 72, 72, 73, 72, 72, 73, 72, 72, 73, 72, 72, 
  73, 72, 72, 73, 72, 72 ]
gap> Lcm(last);
5256
gap> u2nu17^5256; # This element has indeed order 2^3*3^2*73 = 5256.
IdentityMapping( Integers )
gap> u2nu18 := Comm(u2,nu^18);
<rcwa permutation of Z with modulus 81>
gap> cycs := ShortCycles(u2nu18,[-100..100],100);;
gap> List(cycs,Length);
[ 21, 22, 22, 22, 21, 22, 22, 21, 22, 22, 21, 22, 21, 22, 22, 21, 22, 
  22, 21, 22, 22, 21, 22 ]
gap> Lcm(last);
462
gap> u2nu18^462; # This is an element of order 2*3*7*11 = 462.
IdentityMapping( Integers )
gap> List([Comm(u2,nu^20),Comm(u2,nu^25),Comm(u2,nu^30)],Order);
[ 29, 9, 15 ]
]]>
</Example>

We observe that our commutators have various different orders, and
that the prime factors of these orders are not all <Q>very small</Q>. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzlikePerms);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:OddNumberOfGens_FiniteOrder">
<Heading>
  An infinite subgroup of CT(GF(2)[x]) with many torsion elements
</Heading>

In this section, we have a look at the following subgroup of CT(GF(2)[x]):

<Example>
<![CDATA[
gap> x := Indeterminate(GF(2));; SetName(x,"x");
gap> R := PolynomialRing(GF(2),1);
GF(2)[x]
gap> a := ClassTransposition(0,x,1,x);;
gap> b := ClassTransposition(0,x^2+1,1,x^2+1);;
gap> c := ClassTransposition(1,x,0,x^2+x);;
gap> G := Group(a,b,c);
<rcwa group over GF(2)[x] with 3 generators>
gap> Display(G);

Rcwa group over GF(2)[x], generated by

[
Rcwa permutation of GF(2)[x]: P -> P + Z(2)^0

Rcwa permutation of GF(2)[x] with modulus x^2+1, of order 2

        /
        | P + 1 if P in 0(x^2+1) U 1(x^2+1)
 P |-> <  P     if P in x(x^2+1) U x+1(x^2+1)
        |
        \


Rcwa permutation of GF(2)[x] with modulus x^2+x, of order 2

        /
        | (x+1)*P + x+1   if P in 1(x)
 P |-> <  (P + x+1)/(x+1) if P in 0(x^2+x)
        | P               if P in x(x^2+x)
        \

]
]]>
</Example>

We can easily find 2 normal subgroups of&nbsp;<C>G</C>:

<Example>
<![CDATA[
gap> N1 := Subgroup(G,[a*b,a*c]);
<rcwa group over GF(2)[x] with 2 generators>
gap> IsNormal(G,N1);
true
gap> Index(G,N1);
2
gap> G/N1;
Group([ (1,2), (1,2), (1,2) ])
gap> N2 := Subgroup(G,[a*b*c,a*c]);;
gap> IsNormal(G,N2);
true
gap> IsSubgroup(N1,N2);
false
]]>
</Example>

Products of even numbers of generators of&nbsp;<C>G</C> may have
infinite order. For example, we have

<Example>
<![CDATA[
gap> Order(a*b);
2
gap> Order(a*c);
infinity
gap> Order(b*c);
infinity
]]>
</Example>

We would like to have a look at orders of products of odd numbers of
generators. In order to restrict our considerations to <Q>essentially
different</Q> products (as far as we can easily do this), we use
the following auxiliary function:

<Listing Type="GAP code">
<![CDATA[
NormedWords := function ( F, lng )

  local  words, gens, tuples, w;

  gens   := GeneratorsOfGroup(F);
  tuples := EnumeratorOfTuples([1..3],lng);
  words  := [];

  for w in tuples do
    if    (w[1] = 1 or not 1 in w)
      and PositionSublist(w,[1,1]) = fail
      and PositionSublist(w,[2,2]) = fail
      and PositionSublist(w,[3,3]) = fail
      and PositionSublist(w,[2,1]) = fail
      and w[1] < w[lng]
      and w{[1,lng]} <> [1,2]
      and (w{[1..3]} = [1,2,3] or PositionSublist(w,[1,2,3]) = fail)
    then Add(words,w); fi;
  od;

  words := List(words,word->Product(List(word,i->gens[i])));
  return words;
end;
]]>
</Listing>

Now let's compute the possible orders of products of 3, 5, 7 or 9 generators:

<Log>
<![CDATA[
gap> F := FreeGroup("a","b","c");;
gap> phi := EpimorphismByGenerators(F,G);
[ a, b, c ] -> 
[ ClassTransposition(0,x,1,x), ClassTransposition(0,x^2+1,1,x^2+1), 
  ClassTransposition(1,x,0,x^2+x) ]
gap> B3 := NormedWords(F,3);
[ a*b*c ]
gap> B3 := List(B3,g->g^phi);
[ <rcwa permutation of GF(2)[x] with modulus x^3+x> ]
gap> List(B3,Order);
[ 20 ]
gap> B5 := NormedWords(F,5);
[ a*b*c*a*c, a*b*c*b*c ]
gap> B5 := List(B5,g->g^phi);
[ <rcwa permutation of GF(2)[x] with modulus x^3+x>, 
  <rcwa permutation of GF(2)[x] with modulus x^4+x^3+x^2+x> ]
gap> List(B5,Order);
[ 12, 12 ]
gap> B7 := NormedWords(F,7);
[ a*b*c*a*c*a*c, a*b*c*a*c*b*c, a*b*c*b*c*a*c, a*b*c*b*c*b*c ]
gap> B7 := List(B7,g->g^phi);
[ <rcwa permutation of GF(2)[x] with modulus x^4+x^3+x^2+x>, 
  <rcwa permutation of GF(2)[x] with modulus x^5+x>, 
  <rcwa permutation of GF(2)[x] with modulus x^4+x^3+x^2+x>, 
  <rcwa permutation of GF(2)[x] with modulus x^5+x> ]
gap> List(B7,Order);
[ 12, 12, 12, 30 ]
gap> B9 := NormedWords(F,9);
[ a*b*c*a*b*c*a*b*c, a*b*c*a*c*a*c*a*c, a*b*c*a*c*a*c*b*c, a*b*c*a*c*b*c*a*c, 
  a*b*c*a*c*b*c*b*c, a*b*c*b*c*a*c*a*c, a*b*c*b*c*a*c*b*c, a*b*c*b*c*b*c*a*c, 
  a*b*c*b*c*b*c*b*c ]
gap> B9 := List(B9,g->g^phi);;
gap> List(B9,Order);
[ 20, 4, 30, 12, 42, 30, 4, 42, 12 ]
]]>
</Log>

Enter <C>AssignGlobals(LoadRCWAExamples().OddNumberOfGens_FiniteOrder);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:AbelianGroupOverPolynomialRing">
<Heading>An abelian rcwa group over a polynomial ring</Heading>

We enter a 2-generated abelian wild rcwa group over GF(4)[<M>x</M>]:

<Example>
<![CDATA[
gap> x := Indeterminate(GF(4),1);; SetName(x,"x");
gap> R := PolynomialRing(GF(4),1);
GF(2^2)[x]
gap> e := One(GF(4));;
gap> p := x^2 + x + e;;    q := x^2 + e;;
gap> r := x^2 + x + Z(4);; s := x^2 + x + Z(4)^2;;
gap> cg := List( AllResidues(R,x^2), pol -> [ p, p * pol mod q, q ] );;
gap> ch := List( AllResidues(R,x^2), pol -> [ r, r * pol mod s, s ] );;
gap> g := RcwaMapping( R, q, cg );
<rcwa mapping of GF(2^2)[x] with modulus x^2+1>
gap> h := RcwaMapping( R, s, ch );
<rcwa mapping of GF(2^2)[x] with modulus x^2+x+Z(2^2)^2>
gap> List([g,h],IsTame);
[ false, false ]
gap> G := Group(g,h);
<rcwa group over GF(2^2)[x] with 2 generators>
gap> IsAbelian(G);
true
gap> IsTame(G);
false
]]>
</Example>

It is easy to see that all orbits on GF(4)[<M>x</M>] under the action
of <C>G</C> are finite. <P/>

Now we compute the action of the group <C>G</C> on one of its orbits, and
make some statistics of the orbits of <C>G</C> containing polynomials of
degree less than&nbsp;4:

<Example>
<![CDATA[
gap> orb := Orbit(G,x^5);
[ x^5, x^5+x^4+x^2+1, x^5+x^3+x^2+Z(2^2)*x+Z(2)^0, x^5+x^3, 
  x^5+x^4+x^3+x^2+Z(2^2)^2*x+Z(2^2)^2, x^5+x, x^5+x^4+x^3, 
  x^5+x^2+Z(2^2)^2*x, x^5+x^4+x^2+x, x^5+x^3+x^2+Z(2^2)^2*x+Z(2)^0, 
  x^5+x^4+Z(2^2)*x+Z(2^2), x^5+x^3+x, x^5+x^4+x^3+x^2+Z(2^2)*x+Z(2^2), 
  x^5+x^4+x^3+x+1, x^5+x^2+Z(2^2)*x, x^5+x^4+Z(2^2)^2*x+Z(2^2)^2 ]
gap> H := Action(G,orb);
Group([ (1,2,4,7,6,9,12,14)(3,5,8,11,10,13,15,16), 
  (1,3,6,10)(2,5,9,13)(4,8,12,15)(7,11,14,16) ])
gap> IsAbelian(H); # check ...
true
gap> IsCyclic(H);  # H, and therefore also G, is not cyclic
false
gap> Exponent(H);
8
gap> Collected(List(ShortOrbits(G,AllResidues(R,x^4),100),Length));
[ [ 1, 4 ], [ 2, 6 ], [ 4, 12 ], [ 8, 24 ] ]
]]>
</Example>

Changing the generators a little changes the structure of the group
and its action on the underlying ring a lot:

<Example>
<![CDATA[
gap> cg[1][2] := cg[1][2] + (x^2 + e) * p * q;;
gap> ch[7][2] := ch[7][2] + x * r * s;;
gap> g := RcwaMapping( R, q, cg );; h := RcwaMapping( R, s, ch );;
gap> G := Group(g,h);
<rcwa group over GF(2^2)[x] with 2 generators>
gap> IsAbelian(G);
false
gap> Support(G);
GF(2^2)[x] \ [ 1, Z(2^2), Z(2^2)^2 ]
gap> orb := Orbit(G,Zero(R));;
gap> Length(orb);
87
gap> StructureDescription(Action(G,orb));
"A87"
gap> Collected(List(orb,DegreeOfLaurentPolynomial));
[ [ -infinity, 1 ], [ 1, 2 ], [ 2, 4 ], [ 3, 16 ], [ 4, 64 ] ]
gap> S := AllResidues(R,x^6);;
gap> orbs := ShortOrbits(G,S,-1:finite);;
gap> List(orbs,Length);
[ 87, 1, 1, 1, 2, 2, 2, 2, 2, 4, 4, 4, 20, 4, 12, 4, 20, 4, 4, 12, 8, 8, 
  48, 48, 16, 8, 8, 56, 8, 88, 8, 8, 8, 400, 16, 48, 16, 16, 16, 80, 16, 
  16, 16, 96, 32, 192, 32, 16, 16, 416, 16, 48, 16, 16, 880, 16, 16, 16, 
  16, 16, 16, 16, 16, 16, 848, 16, 16, 32, 16, 16, 16, 16, 16, 16, 16 ]
gap> Position(last,880);
55
gap> Set(orbs[55],DegreeOfLaurentPolynomial); # all elm's have same degree
[ 5 ]
gap> H := Action(G,orbs[55]);;
gap> IsPrimitive(H,MovedPoints(H));
false
gap> List(Blocks(H,MovedPoints(H)),Length);
[ 110, 110, 110, 110, 110, 110, 110, 110 ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().AbelianGroupOverPolynomialRing);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:CheckingForSolvability">
<Heading>Checking for solvability</Heading>

Presently there is no general method available for testing wild rcwa groups
for solvability. However, sometimes the question for solvability can be
answered anyway.
In the example below, the idea is to find a subgroup&nbsp;<A>U</A> which acts
on a finite set&nbsp;<A>S</A> of integers, and which induces on&nbsp;<A>S</A>
a non-solvable finite permutation group:

<Example>
<![CDATA[
gap> a := RcwaMapping([[3,0,2],[3, 1,4],[3,0,2],[3,-1,4]]);;
gap> b := RcwaMapping([[3,0,2],[3,13,4],[3,0,2],[3,-1,4]]);;
gap> G := Group(a,b);;
gap> ShortOrbits(Group(Comm(a,b)),[-10..10],100);
[ [ -10 ], [ -9 ], [ -30, -21, -14, -13, -11, -8 ], [ -7 ], [ -6 ], 
  [ -12, -5, -4, -3, -2, 1 ], [ -1 ], [ 0 ], [ 2 ], [ 3 ], 
  [ 4, 5, 6, 7, 10, 15 ], [ 8 ], [ 9 ] ]
gap> S := [ 4, 5, 6, 7, 10, 15 ];;
gap> Cycle(Comm(a,b),4);
[ 4, 7, 10, 15, 5, 6 ]
gap> elm := RepresentativeAction(G,S,Permuted(S,(1,4)),OnTuples);
<rcwa permutation of Z with modulus 81>
gap> List(S,n->n^elm);
[ 7, 5, 6, 4, 10, 15 ]
gap> U := Group(Comm(a,b),elm);
<rcwa group over Z with 2 generators>
gap> Action(U,S);
Group([ (1,4,5,6,2,3), (1,4) ])
gap> IsNaturalSymmetricGroup(last);
true
]]>
</Example>

Thus the subgroup <A>U</A> induces on <A>S</A> a natural symmetric group of
degree&nbsp;6. Therefore the group&nbsp;<A>G</A> is not solvable.
We conclude this example by factoring the group element <A>elm</A> into
generators:

<Example>
<![CDATA[
gap> F := FreeGroup("a","b");
<free group on the generators [ a, b ]>
gap> RepresentativeActionPreImage(G,S,Permuted(S,(1,4)),OnTuples,F);
a^-2*b^-2*a*b*a^-1*b*a*b^-2*a
gap> a^-2*b^-2*a*b*a^-1*b*a*b^-2*a = elm;
true
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().CheckingForSolvability);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:LocalExample">
<Heading>Some examples over (semi)localizations of the integers</Heading>

We start with something one can observe when trying to <Q>transfer</Q> an
rcwa mapping from the ring of integers to one of its localizations:

<Example>
<![CDATA[
gap> a := RcwaMapping([[3,0,2],[3,1,4],[3,0,2],[3,-1,4]]);;
gap> IsBijective(a);
true
gap> a2 := LocalizedRcwaMapping(a,2);
<rcwa mapping of Z_( 2 ) with modulus 4>
gap> IsSurjective(a2); # As expected
true
gap> IsInjective(a2); # Why not??
false
gap> 0^a2;
0
gap> (1/3)^a2; # That's the reason!
0
]]>
</Example>

The above can also be explained easily by pointing out that the
modulus of the inverse of <C>a</C> is&nbsp;3, and that 3 is a unit
of&nbsp;<M>&ZZ;_{(2)}</M>.
Moving to <M>&ZZ;_{(2,3)}</M> solves this problem:

<Example>
<![CDATA[
gap> a23 := SemilocalizedRcwaMapping(a,[2,3]);
<rcwa mapping of Z_( 2, 3 ) with modulus 4>
gap> IsBijective(a23);
true
]]>
</Example>

We get additional finite cycles, e.g.:

<Example>
<![CDATA[
gap> List(ShortOrbits(Group(a23),[0..50]/5,50),orb->Cycle(a23,orb[1]));
[ [ 0 ], [ 1/5, 2/5, 3/5 ], 
  [ 4/5, 6/5, 9/5, 8/5, 12/5, 18/5, 27/5, 19/5, 13/5, 11/5, 7/5 ], 
  [ 1 ], [ 2, 3 ], [ 14/5, 21/5, 17/5 ], 
  [ 16/5, 24/5, 36/5, 54/5, 81/5, 62/5, 93/5, 71/5, 52/5, 78/5, 117/5, 
      89/5, 68/5, 102/5, 153/5, 116/5, 174/5, 261/5, 197/5, 149/5, 
      113/5, 86/5, 129/5, 98/5, 147/5, 109/5, 83/5, 61/5, 47/5, 34/5, 
      51/5, 37/5, 29/5, 23/5 ], [ 4, 6, 9, 7, 5 ] ]
gap> List(last,Length);
[ 1, 3, 11, 1, 2, 3, 34, 5 ]
gap> List(ShortOrbits(Group(a23),[0..50]/7,50),orb->Cycle(a23,orb[1]));
[ [ 0 ], [ -1/7, 1/7 ], [ 2/7, 3/7, 4/7, 6/7, 9/7, 5/7 ], [ 1 ], 
  [ 2, 3 ], [ 4, 6, 9, 7, 5 ] ]
gap> List(last,Length);
[ 1, 2, 6, 1, 2, 5 ]
]]>
</Example>

However the structure of a group with prime set <M>\mathbb{P}</M>
remains invariant under the <Q>transfer</Q> from &ZZ;
to&nbsp;<M>&ZZ;_{(\mathbb{P})}</M>. <P/>

<Q>Transferring</Q> a non-invertible rcwa mapping from the ring of integers
to some of its (semi)localizations can also turn it into an invertible one:

<Example>
<![CDATA[
gap> v := RcwaMapping([[6,0,1],[1,-7,2],[6,0,1],[1,-1,1],
>                      [6,0,1],[1, 1,2],[6,0,1],[1,-1,1]]);;
gap> Display(v);

Rcwa mapping of Z with modulus 8

        /
        | 6n      if n in 0(2)
        | n-1     if n in 3(4)
 n |-> <  (n-7)/2 if n in 1(8)
        | (n+1)/2 if n in 5(8)
        |
        \

gap> IsInjective(v);
true
gap> IsSurjective(v);
false
gap> Image(v);
Z \ 4(12) U 8(12)
gap> Difference(Integers,last);
4(12) U 8(12)
gap> v2 := LocalizedRcwaMapping(v,2);
<rcwa mapping of Z_( 2 ) with modulus 8>
gap> IsBijective(v2);
true
gap> Display(v2^-1);

Rcwa permutation of Z_( 2 ) with modulus 4

        /
        | 1/3 n / 2 if n in 0(4)
        | 2 n + 7   if n in 1(4)
 n |-> <  n + 1     if n in 2(4)
        | 2 n - 1   if n in 3(4)
        |
        \

gap> S := ResidueClass(Z_pi(2),2,0);; l := [S];;
gap> for i in [1..10] do Add(l,l[Length(l)]^v2); od;
gap> l; # Visibly v2 is wild ...
[ 0(2), 0(4), 0(8), 0(16), 0(32), 0(64), 0(128), 0(256), 0(512),
  0(1024), 0(2048) ]
gap> w2 := RcwaMapping(Z_pi(2),[[1,0,2],[2,-1,1],[1,1,1],[2,-1,1]]);;
gap> v2w2 := Comm(v2,w2);; v2w2^-1;;
gap> Display(v2w2);

Rcwa permutation of Z_( 2 ) with modulus 8

        /
        | 3 n   if n in 2(4)
        | n + 4 if n in 1(8)
 n |-> <  n - 4 if n in 5(8)
        | n     if n in 0(4) U 3(4)
        |
        \
]]>
</Example>

Again, viewed as an rcwa mapping of the integers the commutator given at
the end of the example would not be surjective. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().Semilocals);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:Twisting257CyclesToModulus32">
<Heading>
  Twisting 257-cycles into an rcwa mapping with modulus 32
</Heading>

We define an rcwa mapping <A>x</A> of order&nbsp;257 with modulus&nbsp;32.
The easiest way to construct such a mapping is to prescribe a transition
graph and then to assign suitable affine mappings to its vertices.

<Example>
<![CDATA[
gap> x_257 := RcwaMapping(
>      [[ 16,  2,  1], [ 16, 18,  1], [  1, 16,  1], [ 16, 18,  1],
>       [  1, 16,  1], [ 16, 18,  1], [  1, 16,  1], [ 16, 18,  1],
>       [  1, 16,  1], [ 16, 18,  1], [  1, 16,  1], [ 16, 18,  1],
>       [  1, 16,  1], [ 16, 18,  1], [  1, 16,  1], [ 16, 18,  1],
>       [  1,  0, 16], [ 16, 18,  1], [  1,-14,  1], [ 16, 18,  1],
>       [  1,-14,  1], [ 16, 18,  1], [  1,-14,  1], [ 16, 18,  1],
>       [  1,-14,  1], [ 16, 18,  1], [  1,-14,  1], [ 16, 18,  1],
>       [  1,-14,  1], [ 16, 18,  1], [  1,-14,  1], [  1,-31,  1]]);;
gap> Order(x_257);; Display(x_257:CycleNotation:=false);

Rcwa permutation of Z with modulus 32, of order 257

        /
        | 16n+18 if n in 1(2) \ 31(32)
        | n+16   if n in 2(32) U 4(32) U 6(32) U 8(32) U 10(32) U 
        |                12(32) U 14(32)
        | n-14   if n in 18(32) U 20(32) U 22(32) U 24(32) U 26(32) U 
 n |-> <                 28(32) U 30(32)
        | 16n+2  if n in 0(32)
        | n/16   if n in 16(32)
        | n-31   if n in 31(32)
        |
        \

gap> Display(x_257);

Rcwa permutation of Z with modulus 32, of order 257

( 0(32), 2(512), 18(512), 4(512), 20(512), 6(512), 22(512), 
  8(512), 24(512), 10(512), 26(512), 12(512), 28(512), 14(512), 
  30(512), 16(512), 1(32), 34(512), 50(512), 36(512), 52(512), 
  38(512), 54(512), 40(512), 56(512), 42(512), 58(512), 44(512), 
  60(512), 46(512), 62(512), 48(512), 3(32), 66(512), 82(512), 
  68(512), 84(512), 70(512), 86(512), 72(512), 88(512), 74(512), 
  90(512), 76(512), 92(512), 78(512), 94(512), 80(512), 5(32), 
  98(512), 114(512), 100(512), 116(512), 102(512), 118(512), 
  104(512), 120(512), 106(512), 122(512), 108(512), 124(512), 
  110(512), 126(512), 112(512), 7(32), 130(512), 146(512), 
  132(512), 148(512), 134(512), 150(512), 136(512), 152(512), 
  138(512), 154(512), 140(512), 156(512), 142(512), 158(512), 
  144(512), 9(32), 162(512), 178(512), 164(512), 180(512), 
  166(512), 182(512), 168(512), 184(512), 170(512), 186(512), 
  172(512), 188(512), 174(512), 190(512), 176(512), 11(32), 
  194(512), 210(512), 196(512), 212(512), 198(512), 214(512), 
  200(512), 216(512), 202(512), 218(512), 204(512), 220(512), 
  206(512), 222(512), 208(512), 13(32), 226(512), 242(512), 
  228(512), 244(512), 230(512), 246(512), 232(512), 248(512), 
  234(512), 250(512), 236(512), 252(512), 238(512), 254(512), 
  240(512), 15(32), 258(512), 274(512), 260(512), 276(512), 
  262(512), 278(512), 264(512), 280(512), 266(512), 282(512), 
  268(512), 284(512), 270(512), 286(512), 272(512), 17(32), 
  290(512), 306(512), 292(512), 308(512), 294(512), 310(512), 
  296(512), 312(512), 298(512), 314(512), 300(512), 316(512), 
  302(512), 318(512), 304(512), 19(32), 322(512), 338(512), 
  324(512), 340(512), 326(512), 342(512), 328(512), 344(512), 
  330(512), 346(512), 332(512), 348(512), 334(512), 350(512), 
  336(512), 21(32), 354(512), 370(512), 356(512), 372(512), 
  358(512), 374(512), 360(512), 376(512), 362(512), 378(512), 
  364(512), 380(512), 366(512), 382(512), 368(512), 23(32), 
  386(512), 402(512), 388(512), 404(512), 390(512), 406(512), 
  392(512), 408(512), 394(512), 410(512), 396(512), 412(512), 
  398(512), 414(512), 400(512), 25(32), 418(512), 434(512), 
  420(512), 436(512), 422(512), 438(512), 424(512), 440(512), 
  426(512), 442(512), 428(512), 444(512), 430(512), 446(512), 
  432(512), 27(32), 450(512), 466(512), 452(512), 468(512), 
  454(512), 470(512), 456(512), 472(512), 458(512), 474(512), 
  460(512), 476(512), 462(512), 478(512), 464(512), 29(32), 
  482(512), 498(512), 484(512), 500(512), 486(512), 502(512), 
  488(512), 504(512), 490(512), 506(512), 492(512), 508(512), 
  494(512), 510(512), 496(512), 31(32) )

gap> Length(Cycle(x_257,0));
257
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().LongCyclesOfPrimeLength);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:ModuliOfPowers">
<Heading> The behaviour of the moduli of powers </Heading>

We give some examples of how the series of the moduli of powers of a given
rcwa mapping of the integers can look like.

<Example>
<![CDATA[
gap> a := RcwaMapping([[3,0,2],[3, 1,4],[3,0,2],[3,-1,4]]);;
gap> List([0..4],i->Modulus(a^i));
[ 1, 4, 16, 64, 256 ]
gap> e1 := RcwaMapping([[1,4,1],[2,0,1],[1,0,2],[2,0,1]]);;
gap> e2 := RcwaMapping([[1,4,1],[2,0,1],[1,0,2],[1,0,1],
>                       [1,4,1],[2,0,1],[1,0,1],[1,0,1]]);;
gap> List([e1,e2],Order);
[ infinity, infinity ]
gap> List([1..20],i->Modulus(e1^i));
[ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4 ]
gap> List([1..20],i->Modulus(e2^i));
[ 8, 4, 8, 4, 8, 4, 8, 4, 8, 4, 8, 4, 8, 4, 8, 4, 8, 4, 8, 4 ]
gap> Display(e2);

Rcwa permutation of Z with modulus 8, of order infinity

        /
        | n+4 if n in 0(4)
        | 2n  if n in 1(4)
 n |-> <  n/2 if n in 2(8)
        | n   if n in 3(4) U 6(8)
        |
        \

gap> e2^2 = Restriction(RcwaMapping([[1,2,1]]),RcwaMapping([[4,0,1]]));
true
gap> g:=RcwaMapping([[2,2,1],[1, 4,1],[1,0,2],[2,2,1],[1,-4,1],[1,-2,1]]);;
gap> h:=RcwaMapping([[2,2,1],[1,-2,1],[1,0,2],[2,2,1],[1,-1,1],[1, 1,1]]);;
gap> List([0..7],i->Modulus(g^i));
[ 1, 6, 12, 12, 12, 12, 6, 1 ]
gap> List([1..18],i->Modulus((g^3*h)^i));
[ 12, 6, 12, 12, 12, 6, 12, 6, 12, 12, 12, 6, 12, 6, 12, 12, 12, 6 ]
gap> u := RcwaMapping([[3,0,5],[9,1,5],[3,-1,5],[9,-2,5],[9,4,5]]);;
gap> List([0..3],i->Modulus(u^i));
[ 1, 5, 25, 125 ]
gap> v6 := RcwaMapping([[-1,2,1],[1,-1,1],[1,-1,1]]);;
gap> List([0..6],i->Modulus(v6^i));
[ 1, 3, 3, 3, 3, 3, 1 ]
gap> w8 := RcwaMapping([[-1,3,1],[1,-1,1],[1,-1,1],[1,-1,1]]);;
gap> List([0..8],i->Modulus(w8^i));
[ 1, 4, 4, 4, 4, 4, 4, 4, 1 ]
gap> z := RcwaMapping([[2,1,1],[1, 1,1],[2,-1,1],[2, -2,1],
>                      [1,6,2],[1, 1,1],[1,-6,2],[2,  5,1],
>                      [1,6,2],[1, 1,1],[1, 1,1],[2, -5,1],
>                      [1,0,1],[1,-4,1],[1, 0,1],[2,-10,1]]);;
gap> IsBijective(z);
true
gap> List([0..25],i->Modulus(z^i));
[ 1, 16, 32, 64, 64, 128, 128, 128, 128, 128, 128, 256, 256, 256, 256, 
  256, 256, 512, 512, 512, 512, 512, 512, 1024, 1024, 1024 ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().ModuliOfPowers);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:CollatzImagesAndPreImages">
<Heading> Images and preimages under the Collatz mapping </Heading>

We have a look at the images of the residue class 1(2)
under powers of the Collatz mapping.

<Example>
<![CDATA[
gap> T := RcwaMapping([[1,0,2],[3,1,2]]);;
gap> S0 := ResidueClass(Integers,2,1);;
gap> S1 := S0^T;
2(3)
gap> S2 := S1^T;
1(3) U 8(9)
gap> S3 := S2^T;
2(3) U 4(9)
gap> S4 := S3^T;
Z \ 0(3) U 5(9)
gap> S5 := S4^T;
Z \ 0(3) U 7(9)
gap> S6 := S5^T;
Z \ 0(3)
gap> S7 := S6^T;
Z \ 0(3)
]]>
</Example>

Thus the image gets stable after applying the mapping <M>T</M> for
the 6th time. Hence <M>T^6</M> maps the residue class 1(2) surjectively
onto the union of the residue classes 1(3) and 2(3), which <M>T</M>
stabilizes setwise.
Now we would like to determine the preimages of 1(3) and 2(3) in 1(2)
under <M>T^6</M>. The residue class 1(2) has to be the disjoint union of
these sets.

<Example>
<![CDATA[
gap> U := Intersection(PreImage(T^6,ResidueClass(Integers,3,1)),S0);
<union of 11 residue classes (mod 64)>
gap> V := Intersection(PreImage(T^6,ResidueClass(Integers,3,2)),S0);
<union of 21 residue classes (mod 64)>
gap> AsUnionOfFewClasses(U);
[ 1(64), 5(64), 7(64), 9(64), 21(64), 23(64), 29(64), 31(64), 49(64), 
  51(64), 59(64) ]
gap> AsUnionOfFewClasses(V);
[ 3(32), 11(32), 13(32), 15(32), 25(32), 17(64), 19(64), 27(64), 33(64), 
  37(64), 39(64), 41(64), 53(64), 55(64), 61(64), 63(64) ]
gap> Union(U,V) = S0 and Intersection(U,V) = [];  # consistency check
true
]]>
</Example>

The images of the residue class 0(3) under powers of&nbsp;<M>T</M> look
as follows:

<Example>
<![CDATA[
gap> S0 := ResidueClass(Integers,3,0);
0(3)
gap> S1 := S0^T;
0(3) U 5(9)
gap> S2 := S1^T;
0(3) U 5(9) U 7(9) U 8(27)
gap> S3 := S2^T;
<union of 20 residue classes (mod 27) (6 classes)>
gap> S4 := S3^T;
<union of 73 residue classes (mod 81)>
gap> S5 := S4^T;
Z \ 10(81) U 37(81)
gap> S6 := S5^T;
Integers
gap> S7 := S6^T;
Integers
]]>
</Example>

Thus every integer is the image of a multiple of&nbsp;3
under&nbsp;<M>T^6</M>. This means that it would be sufficient to prove the
<M>3n+1</M> conjecture for multiples of&nbsp;3.
We can obtain the corresponding result for multiples of&nbsp;5 as follows:

<Example>
<![CDATA[
gap> S := [ResidueClass(Integers,5,0)];
[ 0(5) ]
gap> for i in [1..12] do Add(S,S[i]^T); od;
gap> for s in S do View(s); Print("\n"); od;
0(5)
0(5) U 8(15)
0(5) U 4(15) U 8(15)
0(5) U 2(15) U 4(15) U 8(15) U 29(45)
<union of 73 residue classes (mod 135)>
<union of 244 residue classes (mod 405)>
<union of 784 residue classes (mod 1215)>
<union of 824 residue classes (mod 1215)>
<union of 2593 residue classes (mod 3645)>
<union of 2647 residue classes (mod 3645)>
<union of 2665 residue classes (mod 3645)>
<union of 2671 residue classes (mod 3645)>
1(3) U 2(3) U 0(15)
gap> Union(S[13],ResidueClass(Integers,3,0));
Integers
gap>  List(S,Si->Float(Density(Si)));
[ 0.2, 0.266667, 0.333333, 0.422222, 0.540741, 0.602469, 0.645267, 
  0.678189, 0.711385, 0.7262, 0.731139, 0.732785, 0.733333 ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzMapping);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:Sigma_T">
<Heading>
  An extension of the Collatz mapping T to a permutation of <M>&ZZ;^2</M>
</Heading>

The Collatz mapping&nbsp;<M>T</M> is surjective, but not injective:

<Example>
<![CDATA[
gap> T := RcwaMapping([[1,0,2],[3,1,2]]);;
gap> Display(T);

Rcwa mapping of Z with modulus 2

        /
        | n/2      if n in 0(2)
 n |-> <  (3n+1)/2 if n in 1(2)
        |
        \

gap> IsInjective(T); IsSurjective(T);
false
true
gap> PreImages(T,2);
[ 1, 4 ]
]]>
</Example>

Often, dealing with rcwa permutations is easier.
Indeed the Collatz mapping&nbsp;<M>T</M> can be extended in natural
ways to permutations of&nbsp;<M>&ZZ;^2</M>. For example, the following
permutation acts on the second coordinate just like&nbsp;<M>T</M>:

<Example>
<![CDATA[
gap> Sigma_T := RcwaMapping( Integers^2, [[1,0],[0,6]],
>                            [[[[2,0],[0,1]],[0,0],2],
>                             [[[4,0],[0,3]],[2,1],2],
>                             [[[2,0],[0,1]],[0,0],2],
>                             [[[4,0],[0,3]],[2,1],2],
>                             [[[4,0],[0,1]],[0,0],2],
>                             [[[4,0],[0,3]],[2,1],2]] );
<rcwa mapping of Z^2 with modulus (1,0)Z+(0,6)Z>
gap> IsBijective(Sigma_T);
true
gap> Display(Sigma_T);

Rcwa permutation of Z^2 with modulus (1,0)Z+(0,6)Z

            /
            | (2m+1,(3n+1)/2) if (m,n) in (0,1)+(1,0)Z+(0,2)Z
            | (m,n/2)         if (m,n) in (0,0)+(1,0)Z+(0,6)Z U 
 (m,n) |-> <                              (0,2)+(1,0)Z+(0,6)Z
            | (2m,n/2)        if (m,n) in (0,4)+(1,0)Z+(0,6)Z
            |
            \

gap> Display(Sigma_T^-1);

Rcwa permutation of Z^2 with modulus (2,0)Z+(0,3)Z

            /
            | (m,2n)             if (m,n) in (0,0)+(1,0)Z+(0,3)Z U 
            |                                (0,1)+(1,0)Z+(0,3)Z
 (m,n) |-> <  (m/2,2n)           if (m,n) in (0,2)+(2,0)Z+(0,3)Z
            | ((m-1)/2,(2n-1)/3) if (m,n) in (1,2)+(2,0)Z+(0,3)Z
            |
            \
]]>
</Example>

Now, the <M>3n+1</M> conjecture is equivalent to the assertion that
the line <M>n=4</M> is a set of representatives for the cycles of
<C>Sigma&uscore;T</C> on the half plane <M>n > 0</M>. <P/>

Let's have a look at a part of a cycle of <C>Sigma&uscore;T</C>:

<Example>
<![CDATA[
gap> Trajectory(Sigma_T,[0,27],75);
[ [ 0, 27 ], [ 1, 41 ], [ 3, 62 ], [ 3, 31 ], [ 7, 47 ], [ 15, 71 ], 
  [ 31, 107 ], [ 63, 161 ], [ 127, 242 ], [ 127, 121 ], [ 255, 182 ], 
  [ 255, 91 ], [ 511, 137 ], [ 1023, 206 ], [ 1023, 103 ], 
  [ 2047, 155 ], [ 4095, 233 ], [ 8191, 350 ], [ 8191, 175 ], 
  [ 16383, 263 ], [ 32767, 395 ], [ 65535, 593 ], [ 131071, 890 ], 
  [ 131071, 445 ], [ 262143, 668 ], [ 262143, 334 ], [ 524286, 167 ], 
  [ 1048573, 251 ], [ 2097147, 377 ], [ 4194295, 566 ], [ 4194295, 283 ], 
  [ 8388591, 425 ], [ 16777183, 638 ], [ 16777183, 319 ], 
  [ 33554367, 479 ], [ 67108735, 719 ], [ 134217471, 1079 ], 
  [ 268434943, 1619 ], [ 536869887, 2429 ], [ 1073739775, 3644 ], 
  [ 1073739775, 1822 ], [ 2147479550, 911 ], [ 4294959101, 1367 ], 
  [ 8589918203, 2051 ], [ 17179836407, 3077 ], [ 34359672815, 4616 ], 
  [ 34359672815, 2308 ], [ 68719345630, 1154 ], [ 68719345630, 577 ], 
  [ 137438691261, 866 ], [ 137438691261, 433 ], [ 274877382523, 650 ], 
  [ 274877382523, 325 ], [ 549754765047, 488 ], [ 549754765047, 244 ], 
  [ 1099509530094, 122 ], [ 1099509530094, 61 ], [ 2199019060189, 92 ], 
  [ 2199019060189, 46 ], [ 4398038120378, 23 ], [ 8796076240757, 35 ], 
  [ 17592152481515, 53 ], [ 35184304963031, 80 ], [ 35184304963031, 40 ], 
  [ 70368609926062, 20 ], [ 70368609926062, 10 ], [ 140737219852124, 5 ], 
  [ 281474439704249, 8 ], [ 281474439704249, 4 ], [ 562948879408498, 2 ], 
  [ 562948879408498, 1 ], [ 1125897758816997, 2 ], 
  [ 1125897758816997, 1 ], [ 2251795517633995, 2 ], 
  [ 2251795517633995, 1 ] ]
gap> Trajectory(Sigma_T^-1,[0,27],20);
[ [ 0, 27 ], [ 0, 54 ], [ 0, 108 ], [ 0, 216 ], [ 0, 432 ], [ 0, 864 ], 
  [ 0, 1728 ], [ 0, 3456 ], [ 0, 6912 ], [ 0, 13824 ], [ 0, 27648 ], 
  [ 0, 55296 ], [ 0, 110592 ], [ 0, 221184 ], [ 0, 442368 ], 
  [ 0, 884736 ], [ 0, 1769472 ], [ 0, 3538944 ], [ 0, 7077888 ], 
  [ 0, 14155776 ] ]
]]>
</Example>

While it seems easy to make conjectures regarding the behaviour of cycles
of <C>Sigma&uscore;T</C>, obtaining results on it is apparently hard.
We observe however that <C>Sigma&uscore;T</C> can be written as a product
of two permutations of&nbsp;<M>&ZZ;^2</M> whose cycles can be described
easily:

<Example>
<![CDATA[
gap> a := RcwaMapping(Integers^2,[[1,0],[0,2]],[[[[4,0],[0,1]],[0, 0],2],
>                                               [[[4,0],[0,1]],[2,-1],2]]);
<rcwa mapping of Z^2 with modulus (1,0)Z+(0,2)Z>
gap> b := a^-1*Sigma_T;
<rcwa permutation of Z^2 with modulus (2,0)Z+(0,3)Z>
gap> Display(a);

Rcwa permutation of Z^2 with modulus (1,0)Z+(0,2)Z

            /
            | (2m,n/2)       if (m,n) in (0,0)+(1,0)Z+(0,2)Z
 (m,n) |-> <  (2m+1,(n-1)/2) if (m,n) in (0,1)+(1,0)Z+(0,2)Z
            |
            \

gap> Display(b);

Rcwa permutation of Z^2 with modulus (2,0)Z+(0,3)Z

            /
            | (m,3n+2) if (m,n) in (1,0)+(2,0)Z+(0,1)Z
            | (m/2,n)  if (m,n) in (0,0)+(2,0)Z+(0,3)Z U 
 (m,n) |-> <                       (0,1)+(2,0)Z+(0,3)Z
            | (m,n)    if (m,n) in (0,2)+(2,0)Z+(0,3)Z
            |
            \
]]>
</Example>

It is easy to see that both <C>a</C> and&nbsp;<C>b</C> have infinite order.
The cycles of&nbsp;<C>a</C> have roughly hyperbolic shape and run, so to
speak, from <M>(0,\pm \infty)</M> to <M>(\pm \infty,0)</M>.
A given cycle contains only finitely many points both of whose coordinates
are nonzero. The fixed points of&nbsp;<C>a</C> are (0,0) and (-1,-1).
We have a look at an example of a cycle of&nbsp;<C>a</C>:

<Example>
<![CDATA[
gap> Trajectory(a,[1000,1000],15);
[ [ 1000, 1000 ], [ 2000, 500 ], [ 4000, 250 ], [ 8000, 125 ], 
  [ 16001, 62 ], [ 32002, 31 ], [ 64005, 15 ], [ 128011, 7 ], 
  [ 256023, 3 ], [ 512047, 1 ], [ 1024095, 0 ], [ 2048190, 0 ], 
  [ 4096380, 0 ], [ 8192760, 0 ], [ 16385520, 0 ] ]
gap> Trajectory(a^-1,[1000,1000],15);
[ [ 1000, 1000 ], [ 500, 2000 ], [ 250, 4000 ], [ 125, 8000 ], 
  [ 62, 16001 ], [ 31, 32002 ], [ 15, 64005 ], [ 7, 128011 ], 
  [ 3, 256023 ], [ 1, 512047 ], [ 0, 1024095 ], [ 0, 2048190 ], 
  [ 0, 4096380 ], [ 0, 8192760 ], [ 0, 16385520 ] ]
]]>
</Example>

It is left as an easy exercise to the reader to find out how the cycles
of&nbsp;<C>b</C> look like. <P/>

Enter <C>AssignGlobals(LoadRCWAExamples().ZxZ);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:GrigorchukGroups">
<Heading>
  Finite quotients of Grigorchuk groups
</Heading>

In this section, we show how to construct finite quotients of the two
infinite periodic groups introduced by Rostislav Grigorchuk
in&nbsp;<Cite Key="Grigorchuk80"/> with the help of &RCWA;.

The first of these, nowadays known as <Q>Grigorchuk group</Q>, is
investigated in an example given on the &GAP; website -- see
<URL>http://www.gap-system.org/Doc/Examples/grigorchuk.html</URL>.

The &RCWA; package permits a simpler and more elegant construction
of the finite quotients of this group: The function <C>TopElement</C>
given on the mentioned webpage gets unnecessary, and the function
<C>SequenceElement</C> can be simplified as follows:

<Listing>
<![CDATA[
SequenceElement := function ( r, level )

  return Permutation(Product(Filtered([1..level-1],k->k mod 3 <> r),
                             k->ClassTransposition(    2^(k-1)-1,2^(k+1),
                                                   2^k+2^(k-1)-1,2^(k+1))),
                     [0..2^level-1]);
end;
]]>
</Listing>

The actual constructors for the generators are modified as follows:

<Listing>
<![CDATA[
a := level -> Permutation(ClassTransposition(0,2,1,2),[0..2^level-1]);
b := level -> SequenceElement(0,level);
c := level -> SequenceElement(2,level);
d := level -> SequenceElement(1,level);
]]>
</Listing>

All computations given on the webpage can now be done just as with the
<Q>original</Q> construction of the quotients of the Grigorchuk group.
In the sequel, we construct finite quotients of the second group introduced
in&nbsp;<Cite Key="Grigorchuk80"/>:

<Example>
<![CDATA[
gap> FourCycle := RcwaMapping((4,5,6,7),[4..7]);
( 0(4), 1(4), 2(4), 3(4) )
gap> GrigorchukGroup2Generator := function ( level )
>      if level = 1 then return FourCycle; else
>        return   Restriction(FourCycle, RcwaMapping([[4,1,1]]))
>               * Restriction(FourCycle, RcwaMapping([[4,3,1]]))
>               * Restriction(GrigorchukGroup2Generator(level-1),
>                             RcwaMapping([[4,0,1]]));
>      fi;
>    end;;
gap> GrigorchukGroup2 := level -> Group(FourCycle,
>                                       GrigorchukGroup2Generator(level));;
]]>
</Example>

We can do similar things as shown in the example on the &GAP; webpage
for the <Q>first</Q> Grigorchuk group:

<Example>
<![CDATA[
gap> G := List([1..4],lev->GrigorchukGroup2(lev)); # The first 4 quotients.
[ <rcwa group over Z with 2 generators>, 
  <rcwa group over Z with 2 generators>, 
  <rcwa group over Z with 2 generators>, 
  <rcwa group over Z with 2 generators> ]
gap> H := List([1..4],lev->Action(G[lev],[0..4^lev-1])); # Isom. perm.-gps.
[ Group([ (1,2,3,4), (1,2,3,4) ]), 
  Group([ (1,2,3,4)(5,6,7,8)(9,10,11,12)(13,14,15,16), 
      (1,5,9,13)(2,6,10,14)(4,8,12,16) ]), 
  <permutation group with 2 generators>, 
  <permutation group with 2 generators> ]
gap> List(H,Size);
[ 4, 1024, 4294967296, 1329227995784915872903807060280344576 ]
gap> List(last,n->Collected(Factors(n)));
[ [ [ 2, 2 ] ], [ [ 2, 10 ] ], [ [ 2, 32 ] ], [ [ 2, 120 ] ] ]
gap> List(H,NilpotencyClassOfGroup);      
[ 1, 6, 14, 40 ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().GrigorchukQuotients);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:ForwardOrbit">
<Heading>
  Forward orbits of a monoid with 2 generators
</Heading>

The <M>3n+1</M> conjecture asserts that the forward orbit of any positive
integer under the Collatz mapping <M>T</M> contains&nbsp;1. In contrast,
it seems likely that <Q>most</Q> trajectories of the two mappings
<Alt Only="LaTeX">
  <Display>
  <![CDATA[T_5^\pm: \ \mathbb{Z} \longrightarrow \mathbb{Z}, \ \ \ \
    n \ \longmapsto \
    \begin{cases}
      \frac{n}{2}        & \text{if} \ n \ \text{even}, \\
      \frac{5n \pm 1}{2} & \text{if} \ n \ \text{odd}
    \end{cases}]]>
  </Display>
</Alt>
<Alt Only="HTML"><![CDATA[<center>
  <img src = "t5pm.png" width = "372" height = "61"
       alt = "T5+/-: Z -> Z, n |-> (n/2 if n even, (5n+/-1)/2 if n odd)" />
</center>]]></Alt>
<Alt Only="Text"><Verb><![CDATA[
                                      /
                                      | n/2          if n even,
          T_5+/-:  Z -> Z,   n  |->  <
                                      | (5n +/- 1)/2 if n odd
                                      \
]]></Verb></Alt>
diverge.
However we can show by means of computation that the forward orbit of any
positive integer under the action of the monoid generated by the two
mappings <M>T_5^-</M> and&nbsp;<M>T_5^+</M> indeed contains&nbsp;1.
First of all, we enter the generators:

<Example>
<![CDATA[
gap> T5m := RcwaMapping([[1,0,2],[5,-1,2]]);;
gap> T5p := RcwaMapping([[1,0,2],[5, 1,2]]);;
]]>
</Example>

We look for a number <M>k</M> such that for any residue class <M>r(2^k)</M>
there is a product&nbsp;<M>f</M> of <M>k</M>&nbsp;mappings <M>T_5^\pm</M>
whose restriction to <M>r(2^k)</M> is given by <M>n \mapsto (an+b)/c</M>
where <M>c>a</M>:

<Example>
<![CDATA[
gap> k := 1;;
gap> repeat
>      maps := List(Tuples([T5m,T5p],k),Product);
>      decr := List(maps,DecreasingOn);
>      decreasable := Union(decr);
>      Print(k,": "); View(decreasable); Print("\n");
>      k := k + 1;
>    until decreasable = Integers;
1: 0(2)
2: 0(4)
3: Z \ 1(8) U 7(8)
4: 0(4) U 3(16) U 6(16) U 10(16) U 13(16)
5: Z \ 7(32) U 25(32)
6: <union of 48 residue classes (mod 64)>
7: Integers
]]>
</Example>

Thus <M>k=7</M> serves our purposes.
To be sure that for any positive integer <M>n</M> our monoid contains
a mapping <M>f</M> such that <M>n^f&lt;n</M>, we still need to check this
condition for <Q>small</Q>&nbsp;<M>n</M>. Since in case <M>c>a</M> we have
<M>(an+b)/c \geq n</M> if only if <M>n \leq b/(c-a)</M>, we only need to
check those <M>n</M> which are not larger than the largest coefficient
<M>b_{r(m)}</M> occurring in any of the products under consideration:

<Example>
<![CDATA[
gap> maxb := Maximum(List(maps,f->Maximum(List(Coefficients(f),t->t[2]))));
25999
gap> small := Filtered([1..maxb],n->ForAll(maps,f->n^f>=n));
[ 1, 7, 9, 11 ]
]]>
</Example>

This means that except of&nbsp;1, only for <M>n \in \{{7,9,11\}}</M> there
is no product of 7 mappings <M>T_5^\pm</M> which maps <M>n</M> to a smaller
integer. We check that also the forward orbits of these three integers
contain&nbsp;1 by successively computing preimages of&nbsp;1: 

<Example>
<![CDATA[
gap> S := [1];; k := 0;;
gap> repeat
>      S := Union(S,PreImage(T5m,S),PreImage(T5p,S));
>      k := k+1;
>    until IsSubset(S,small);
gap> k;
17
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().CollatzMapping);</C>
in order to assign the global variables defined in this section.

</Section>

<!-- #################################################################### -->

<Section Label="sec:F2andPSL2Z">
<Heading>
  The free group of rank 2 and the modular group PSL(2,&ZZ;)
</Heading>

The free group of rank&nbsp;2 embeds into RCWA(&ZZ;) -- in fact it embeds
even in the subgroup which is generated by all class transpositions.
An explicit embedding can be constructed by transferring the construction
of the so-called <Q>Schottky groups</Q> (cf.&nbsp;<Cite Key="LaHarpe00"/>,
page&nbsp;27) from PSL(2,&CC;) to RCWA(&ZZ;)
(we use the notation from the cited book):

<Example>
<![CDATA[
gap> D := AllResidueClassesModulo(4);
[ 0(4), 1(4), 2(4), 3(4) ]
gap> gamma1 := RepresentativeAction(RCWA(Integers),
>                                   Difference(Integers,D[1]),D[2]);;
gap> gamma2 := RepresentativeAction(RCWA(Integers),
>                                   Difference(Integers,D[3]),D[4]);;
gap> F2 := Group(gamma1,gamma2);
<rcwa group over Z with 2 generators>
]]>
</Example>

We can do some checks:

<Example>
<![CDATA[
gap> X1 := Union(D{[1,2]});; X2 := Union(D{[3,4]});;
gap>     IsSubset(X1,X2^gamma1) and IsSubset(X1,X2^(gamma1^-1))
>    and IsSubset(X2,X1^gamma2) and IsSubset(X2,X1^(gamma2^-1));
true
]]>
</Example>

The generators are products of 3 class transpositions, each:

<Example>
<![CDATA[
gap> Factorization(gamma1);
[ ( 0(2), 1(2) ), ( 3(4), 5(8) ), ( 0(2), 1(8) ) ]
gap> Factorization(gamma2);
[ ( 0(2), 1(2) ), ( 1(4), 7(8) ), ( 0(2), 3(8) ) ]
]]>
</Example>

The above construction is used by <Ref Attr="IsomorphismRcwaGroup"
Label="for a group"/> to embed free groups of any rank&nbsp;<M>\geq 2</M>.
<P/>

We give another only slightly different representation of the free group
of rank&nbsp;2. We verify that it really is one by applying the
so-called <E>Table-Tennis Lemma</E> (see e.g.&nbsp;<Cite Key="LaHarpe00"/>,
Section&nbsp;II.B.) to the infinite cyclic groups generated by the two
generators and to the same two sets <C>X1</C> and <C>X2</C> as above:

<Example>
<![CDATA[
gap> r1 := ClassTransposition(0,2,1,2)*ClassTransposition(0,2,1,4);;
gap> r2 := ClassTransposition(0,2,1,2)*ClassTransposition(0,2,3,4);;
gap> F2 := Group(r1^2,r2^2);;
gap> List(GeneratorsOfGroup(F2),IsTame);
[ false, false ]
gap>     IsSubset(X1,X2^F2.1) and IsSubset(X1,X2^(F2.1^-1))
>    and IsSubset(X2,X1^F2.2) and IsSubset(X2,X1^(F2.2^-1));
true
gap> [Sources(r1),Sinks(r1),Loops(r1)]; # compare with X1
[ [ 0(4) ], [ 1(4) ], [ 0(4), 1(4) ] ]
gap> [Sources(r2),Sinks(r2),Loops(r2)]; # compare with X2
[ [ 2(4) ], [ 3(4) ], [ 2(4), 3(4) ] ]
gap>    IsSubset(X1,Union(Sinks(r1))) and IsSubset(X1,Union(Sinks(r1^-1)))
>   and IsSubset(X2,Union(Sinks(r2))) and IsSubset(X2,Union(Sinks(r2^-1)));
true
gap> IsSubset(Union(Sinks(r1)),X2^F2.1) and
>    IsSubset(Union(Sinks(r1^-1)),X2^(F2.1^-1));
true
gap> IsSubset(Union(Sinks(r2)),X1^F2.2) and
>    IsSubset(Union(Sinks(r2^-1)),X1^(F2.2^-1));
true
]]>
</Example>

Drawing the transition graphs of <C>r1</C> and <C>r2</C> for modulus&nbsp;4
may help to understand what is actually done in this calculation.
It is easy to see that the group generated by <C>r1</C> and&nbsp;<C>r2</C>
is <E>not</E> free:

<Example>
<![CDATA[
gap> Order(r1/r2);
3
]]>
</Example>

The modular group PSL(2,&ZZ;) embeds into CT(&ZZ;) as well.
We give an embedding, and check that it really is one by applying
the Table Tennis Lemma as above:

<Example>
<![CDATA[
gap> PSL2Z := 
>      Group(ClassTransposition(0,3,1,3) * ClassTransposition(0,3,2,3),
>            ClassTransposition(1,3,0,6) * ClassTransposition(2,3,3,6));;
gap> List(GeneratorsOfGroup(PSL2Z),Order);
[ 3, 2 ]
gap> X1 := Difference(Integers,ResidueClass(0,3));
Z \ 0(3)
gap> X2 := ResidueClass(0,3);
0(3)
gap> IsSubset(X1,X2^PSL2Z.1) and IsSubset(X1,X2^(PSL2Z.1^2));
true
gap> IsSubset(X2,X1^PSL2Z.2);
true
]]>
</Example>

A slightly different representation of PSL(2,&ZZ;) can be obtained by using
&RCWA;'s general method for <C>IsomorphismRcwaGroup</C> for free products of
finite groups:

<Example>
<![CDATA[
gap> G := Image(IsomorphismRcwaGroup(FreeProduct(CyclicGroup(3),
>                                                CyclicGroup(2))));
<wild rcwa group over Z with 2 generators>
gap> List(GeneratorsOfGroup(G),Factorization);
[ [ ( 0(4), 2(4) ), ( 1(2), 0(4) ) ], [ ( 0(2), 1(2) ) ] ]
]]>
</Example>

Enter <C>AssignGlobals(LoadRCWAExamples().F2_PSL2Z);</C>
in order to assign the global variables defined in this section.

<Alt Only="HTML">&nbsp;</Alt>

</Section>

<!-- #################################################################### -->

</Chapter>

<!-- #################################################################### -->