<!-- %W hamilcyc.xml GAP 4 package CTblLib Thomas Breuer -->
<Chapter Label="chap:hamilcyc">
<Heading>&GAP; Computations Concerning Hamiltonian Cycles in the Generating Graphs of Finite Groups</Heading>
Date: April 24th, 2012
<P/>
This is a collection of examples showing how
the &GAP; system <Cite Key="GAP"/>
can be used to compute information about the generating graphs of
finite groups.
It includes all examples that were needed for the computational results
in <Cite Key="GMN"/>.
<P/>
The purpose of this writeup is twofold.
On the one hand, the computations are documented this way.
On the other hand, the &GAP; code shown for the examples can be used as
test input for automatic checking of the data and the functions used.
<P/>
A first version of this document, which was based on &GAP; 4.4.12,
is available in the arXiv at <URL>http://arxiv.org/abs/0911.5589v1</URL>
since November 2009.
The differences between this file and the current document are as follows.
<P/>
<List>
<Item>
The format of the &GAP; output was adjusted to the changed behaviour
of &GAP; 4.5.
</Item>
<Item>
The records returned by
<Ref Func="IsomorphismTypeInfoFiniteSimpleGroup" BookName="ref"/>
contain a component <C>"shortname"</C> since &GAP; 4.11.
</Item>
<Item>
The lower bounds computed for the sporadic simple Monster group
have been improved in three steps.
<P/>
First, the existence of exactly one class of maximal subgroups of the type
<M>&PSL;(2, 41)</M> (see <Cite Key="NW12"/>)
and the nonexistence of maximal subgroups with socle <M>&PSL;(2, 27)</M>
(see <Cite Key="Wil10"/>)
have been incorporated.
<P/>
Second, the classification of classes of maximal subgroups of the Monster
has been completed in <Cite Key="DLP25"/>.
As a consequence, the nonexistence of maximal subgroups with socle
Sz<M>(8)</M> and <M>&PSU;(3, 8)</M>
and the existence of exactly one class of maximal subgroups with the
isomorphism types <M>&PSL;(2, 13).2</M> and <M>&PSU;(3, 4).4</M>
have been proved,
and the proof of the nonexistence of the previously claimed subgroups
of the type <M>L_2(59)</M> yields that the Monster has maximal subgroups
of the type <M>59:29</M>.
Moreover, meanwhile also all class fusions of the maximal subgroups are
known;
previously, we got only candidates for some primitive permutation characters.
<P/>
Third, all character tables of maximal subgroups of the Monster group
and their class fusions are available since version 1.3.10 of
&CTblLib;,
hence no special treatment for the Monster group is needed anymore
in order to compute its primitive permutation characters.
In particular, the data file <F>data/prim_perm_M.json</F> of &CTblLib;
is no longer needed.
In earlier versions, that file had been used to collect information about
primitive permutation characters for which the character table was not yet
available.
<P/>
See Section <Ref Subsect="Monster"/> for comments on earlier versions,
and for the bounds that had been computed from the partial information
that was available at that time.
</Item>
</List>
The purpose of this note is to document the &GAP; computations
that were carried out in order to obtain the computational results
in <Cite Key="GMN"/>.
<P/>
In order to keep this note self-contained,
we first describe the theory needed,
in Section <Ref Subsect="background"/>.
The translation of the relevant formulae into &GAP; functions
can be found in Section <Ref Subsect="functions"/>.
Then Section <Ref Subsect="chartheor"/> describes the computations
that only require (ordinary) character tables in the
&GAP; Character Table Library <Cite Key="CTblLib"/>.
Computations using also the groups are shown in
Section <Ref Subsect="grouptheor"/>.
<P/>
The examples use the &GAP; Character Table Library
and the &GAP; Library of Tables of Marks,
so we first load these packages in the required versions.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#W hamilcyc.g Hamiltonian cycles in finite groups Thomas Breuer
##
#Y Copyright (C) 2009, Lehrstuhl D für Mathematik, RWTH Aachen, Germany
##
#############################################################################
##
## The examples use the GAP Character Table Library
## and the GAP Library of Tables of Marks,
## so we first load these packages in the required versions.
##
]]></Ignore>
<Example><![CDATA[
gap> if not CompareVersionNumbers( GAPInfo.Version, "4.5" ) then
> Error( "need GAP in version at least 4.5" );
> fi;
gap> LoadPackage( "ctbllib", "1.2", false );
true
gap> LoadPackage( "tomlib", "1.1.1", false );
true
]]></Example>
Let <M>G</M> be a finite noncyclic group
and denote by <M>G^{\times}</M> the set of nonidentity elements in <M>G</M>.
We define the <E>generating graph</E> <M>\Gamma(G)</M> as the undirected graph
on the vertex set <M>G^{\times}</M> by joining two elements
<M>x, y \in G^{\times}</M>
by an edge if and only if <M>\langle x, y \rangle = G</M> holds.
For <M>x \in G^{\times}</M>, the <E>vertex degree</E> <M>d(\Gamma, x)</M> is
<M>|\{ y \in G^{\times}; \langle x, y \rangle = G \}|</M>.
The <E>closure</E> <M>&cl;(\Gamma)</M> of the graph <M>\Gamma</M> with
<M>m</M> vertices is defined as the graph with the same vertex set as
<M>\Gamma</M>,
where the vertices <M>x, y</M> are joined by an edge
if they are joined by an edge in <M>\Gamma</M>
or if <M>d(\Gamma, x) + d(\Gamma, y) \geq m</M>.
We denote iterated closures by
<M>&cl;^{(i)}(\Gamma) = &cl;(&cl;^{(i-1)}(\Gamma))</M>,
where <M>&cl;^{(0)}(\Gamma) = \Gamma</M>.
<P/>
In the following, we will show that the generating graphs of the following
groups contain a Hamiltonian cycle:
<List>
<Item>
Nonabelian simple groups of orders at most <M>10^7</M>,
</Item>
<Item>
groups <M>G</M> containing a unique minimal normal subgroup <M>N</M>
such that <M>N</M> has order at most <M>10^6</M>, <M>N</M> is nonsolvable,
and <M>G/N</M> is cyclic, <!-- this covers in particular almost simple groups such that <M>G/</M>Soc<M>(G)</M> is cyclic and
Soc<M>(G)</M> has order at most <M>10^6</M> -->
</Item>
<Item>
sporadic simple groups and their automorphism groups.
</Item>
</List>
<P/>
Clearly the condition that <M>G/N</M> is cyclic for all nontrivial normal
subgroups <M>N</M> of <M>G</M> is necessary for <M>\Gamma(G)</M> being
connected,
and <Cite Where="Conjecture 1.6" Key="GMN"/> states
that this condition is also sufficient.
By <Cite Where="Proposition 1.1" Key="GMN"/>,
this conjecture is true for all solvable groups,
and the second entry in the above list implies that this conjecture holds
for all nonsolvable groups of order up to <M>10^6</M>.
<P/>
The question whether a graph <M>\Gamma</M> contains a Hamiltonian cycle
(i. e., a closed path in <M>\Gamma</M> that visits each vertex
exactly once)
can be answered using the following sufficient criteria
(see <Cite Key="GMN"/>).
Let <M>d_1 \leq d_2 \leq \cdots \leq d_m</M> be the vertex degrees
in <M>\Gamma</M>.
<P/>
<List>
<Mark>Pósa's criterion:
<Item>
If <M>d_k \geq k+1</M> holds for <M>1 \leq k < m/2</M>
then <M>\Gamma</M> contains a Hamiltonian cycle.
</Item>
<Mark>Chvátal's criterion:
<Item>
If <M>d_k \geq k+1</M> or <M>d_{m-k} \geq m-k</M> holds for
<M>1 \leq k < m/2</M>
then <M>\Gamma</M> contains a Hamiltonian cycle.
</Item>
<Mark>Closure criterion:</Mark>
<Item>
A graph contains a Hamiltonian cycle if and only if its closure
contains a Hamiltonian cycle.
</Item>
</List>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="sect:Character-Theoretic Lower Bounds for Vertex Degrees">
<Heading>Character-Theoretic Lower Bounds for Vertex Degrees</Heading>
Using character-theoretic methods similar to those used to obtain the
results in <Cite Key="BGK"/>
(the computations for that paper are shown in <Cite Key="ProbGenArxiv"/>),
we can compute lower bounds for the vertex degrees in generating graphs,
as follows.
<P/>
Let <M>R</M> be a set of representatives of conjugacy classes of nonidentity
elements in <M>G</M>,
fix <M>s \in G^{\times}</M>,
let <M>&M;(G,s)</M> denote the set of those maximal subgroups of <M>G</M>
that contain <M>s</M>,
let <M>&M;(G,s)/\sim</M> denote a set of representatives in <M>&M;(G,s)</M>
w. r. t. conjugacy in <M>G</M>.
For a subgroup <M>M</M> of <M>G</M>,
the <E>permutation character</E> <M>1_M^G</M> is defined by
where <M>g^G = \{ g^x; x \in G \}</M>, with <M>g^x = x^{-1} g x</M>,
denotes the conjugacy class of <M>g</M> in <M>G</M>.
So we have <M>1_M^G(1) = |G|/|M|</M>
and thus <M>|g^G \cap M| = |g^G| \cdot 1_M^G(g) / 1_M^G(1)</M>.
<P/>
Doubly counting the set <M>\{ (s^x, M^y); x, y \in G, s^x \in M^y \}</M>
yields
<M>|M^G| \cdot |s^G \cap M| = |s^G| \cdot |\{ M^x; x \in G, s \in M^x \}|</M>
and thus
<M>|\{ M^x; x \in G, s \in M^x \}| =
|M^G| \cdot 1_M^G(s) / 1_M^G(1) \leq 1_M^G(s)</M>.
(If <M>M</M> is a <E>maximal</E> subgroup of <M>G</M>
then either <M>M</M> is normal in <M>G</M> or self-normalizing,
and in the latter case the inequality is in fact an equality.)
<P/>
Let <M>\Pi</M> denote the multiset of <E>primitive</E> permutation characters
of <M>G</M>, i. e.,
of the permutation characters <M>1_M^G</M> where <M>M</M> ranges over
representatives of the conjugacy classes of maximal subgroups of <M>G</M>.
and <M>d(s, g^G):= |\{ x \in g^G; \langle s, x \rangle = G \}|</M>,
the contribution of the class <M>g^G</M> to the vertex degree of <M>s</M>.
Then we have <M>d(\Gamma(G), s) = \sum_{{x \in R}} d(s, x^G)</M> and
So <M>\delta(s):= \sum_{x \in R} \delta(s, x^G)</M> is a lower bound
for the vertex degree of <M>s</M>; this bound can be computed if <M>\Pi</M> is known.
<P/>
For computing the vertex degrees of the iterated closures of <M>\Gamma(G)</M>,
we define <M>d^{(0)}(s, g^G):= d(s, g^G)</M> and
and <M>\delta^{(i)}(s):= \sum_{{x \in R}} \delta^{(i)}(s, x^G)</M>,
a lower bound for <M>d(&cl;^{(i)}(\Gamma(G)), s)</M>
that can be computed if <M>\Pi</M> is known.
</Subsection>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="critcheck">
<Heading>Checking the Criteria</Heading>
Let us assume that we know lower bounds <M>\beta(s)</M> for the vertex degrees
<M>d(&cl;^{(i)}(\Gamma(G)), s)</M>, for some fixed <M>i</M>,
and let us choose representatives <M>s_1, s_2, \ldots, s_l</M> of the nonidentity
conjugacy classes of <M>G</M>
such that <M>\beta(s_1) \leq \beta(s_2) \leq \cdots \leq \beta(s_l)</M> holds.
Let <M>c_k = |s_k^G|</M> be the class lengths of these representatives.
<P/>
Then the first <M>c_1</M> vertex degrees, ordered by increasing size,
are larger than or equal to <M>\beta(s_1)</M>,
the next <M>c_2</M> vertex degrees are larger than or equal to <M>\beta(s_2)</M>,
and so on.
<P/>
Then the set of indices in the <M>k</M>-th nonidentity class of <M>G</M>
for which Pósa's criterion is not guaranteed by the given bounds is
(Note that the generating graph has <M>m = |G|-1</M> vertices,
and that <M>x < m/2</M> is equivalent to
<M>x \leq \left\lfloor |G|/2 \right\rfloor - 1</M>.)
<P/>
The generating graph <M>\Gamma(G)</M> satisfies Pósa's criterion
if all these intervals are empty,
i. e., if <M>L_k > U_k</M> holds for <M>1 \leq k \leq l</M>.
<P/>
The set of indices for which Chvátal's criterion is not guaranteed
is the intersection of
with the set of indices for which Pósa's criterion is not guaranteed.
<P/>
Analogously to the above considerations,
the set of indices <M>m-x</M> in the former set for which Chvátal's criterion
is not guaranteed by the given bounds
and such that <M>x</M> is an index in the <M>k</M>-th nonidentity class of <M>G</M>
is
The generating graph <M>\Gamma(G)</M> satisfies Chvátal's criterion if
the union of the intervals
<M>\{ L^{\prime}_k, L^{\prime}_k + 1, \ldots, U^{\prime}_k \}</M>,
for <M>1 \leq k \leq l</M>
is disjoint to the union of the intervals <M>\{ L_k, L_k + 1, \ldots, U_k \}</M>,
for <M>1 \leq k \leq l</M>.
</Subsection>
</Section>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Section Label="functions">
<Heading>&GAP; Functions for the Computations</Heading>
We describe two approaches to compute, for a given group <M>G</M>,
vertex degrees for the generating graph of <M>G</M> or lower bounds for them,
by calculating exact vertex degrees from <M>G</M> itself
(see Section <Ref Subsect="groups"/>)
or by deriving lower bounds for the vertex degrees using just
character-theoretic information about <M>G</M> and its subgroups
(see Section <Ref Subsect="characters"/>).
Finally, Section <Ref Subsect="clos"/> deals with deriving lower bounds
of vertex degrees of iterated closures.
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="groups">
<Heading>Computing Vertex Degrees from the Group</Heading>
In this section,
the task is to compute the vertex degrees <M>d(s,g^G)</M> using explicit
computations with the group <M>G</M>.
<P/>
The function <C>IsGeneratorsOfTransPermGroup</C> checks whether the permutations
in the list <C>list</C> generate the permutation group <C>G</C>,
<E>provided that</E> <C>G</C> is transitive on its moved points.
(Note that testing the necessary condition that the elements in <C>list</C>
generate a transitive group is usually much faster than testing generation.)
This function has been used already in <Cite Key="ProbGenArxiv"/>.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F IsGeneratorsOfTransPermGroup( <G>, <list> )
##
## Let <G> be a finite group that acts *transitively* on its moved points,
## and <list> a list of elements in <G>.
##
## 'IsGeneratorsOfTransPermGroup' returns 'true' if the permutations in
## <list> generate <G>, and 'false' otherwise.
## The main point is that the return value 'true' requires the group
## generated by 'list' to be transitive, and the check for transitivity
## is much cheaper than the test whether this group is equal to 'G'.
##
## (The same function is used also in another examples file,
## do not use 'BindGlobal'.)
##
]]></Ignore>
<Example><![CDATA[
gap> IsGeneratorsOfTransPermGroup:= function( G, list )
> local S;
>
> if not IsTransitive( G ) then
> Error( " must be transitive on its moved points" );
> fi;
> S:= SubgroupNC( G, list );
>
> return IsTransitive( S, MovedPoints( G ) )
> and Size( S ) = Size( G );
> end;;
]]></Example>
<P/>
The function <C>VertexDegreesGeneratingGraph</C> takes a
<E>transitive</E> permutation group <M>G</M>
(in order to be allowed to use <C>IsGeneratorsOfTransPermGroup</C>),
the list <C>classes</C> of conjugacy classes of <M>G</M>
(in order to prescribe an ordering of the classes),
and a list <C>normalsubgroups</C> of proper normal subgroups of <M>G</M>,
and returns the matrix <M>[ d(s, g^G) ]_{s, g}</M> of vertex degrees,
with rows and columns indexed by nonidentity class representatives
ordered as in the list <C>classes</C>.
(The class containing the identity element may be contained in <C>classes</C>.)
<P/>
The following criteria are used in this function.
<P/>
<List>
<Item>
The function tests the (non)generation only for representatives of
<M>C_G(g)</M>-<M>C_G(s)</M>-double cosets,
where <M>C_G(g):= \{ x \in G; g x = x g \}</M>
denotes the centralizer of <M>g</M> in <M>G</M>.
Note that for <M>c_1 \in C_G(g)</M>, <M>c_2 \in C_G(s)</M>,
and a representative <M>r \in G</M>, we have
<M>\langle s, g^{c_1 r c_2} \rangle = \langle s, g^r \rangle^{c_2}</M>.
If <M>\langle s, g^r \rangle = G</M> then
the double coset <M>D = C_G(g) r C_G(s)</M> contributes <M>|D|/|C_G(g)|</M>
to the vertex degree <M>d(s, g^G)</M>,
otherwise the contribution is zero. <!-- % explicitly: --> <!-- % \begin{eqnarray*} --> <!-- % d(s,g^G) & = & |\{ x \in g^G, \langle s, x \rangle = G \}| \\ --> <!-- % & = & |\{ x \in G, \langle s, g^x \rangle = G \}| / |C_G(g)| \\ --> <!-- % & = & \sum_{r \in R} |\{ x \in C_G(g) r C_G(s), --> <!-- % \langle s, g^r \rangle = G \}| / |C_G(g)| \\ --> <!-- % & = & \sum_{r} |C_G(g) r C_G(s)| / |C_G(g)|, --> <!-- % \end{eqnarray*} --> <!-- % where <M>R</M> is a set of double coset representatives, i. e., --> <!-- % \[ --> <!-- % G = \bigcup_{r \in R} C_G(g) r C_G(s), --> <!-- % \] --> <!-- % and the last sum runs over all those representatives <M>r</M> such that --> <!-- % <M>\langle s, g^r \rangle = G</M> holds. -->
</Item>
<Item>
We have <M>d(s, g^G) \cdot |C_G(g)| = d(g, s^G) \cdot |C_G(s)|</M>.
(To see this,
either establish a bijection of the above double cosets,
or doubly count the edges between elements of the conjugacy classes
of <M>s</M> and <M>g</M>.)
</Item>
<Item>
If <M>\langle s_1 \rangle = \langle s_2 \rangle</M> and
<M>\langle g_1 \rangle = \langle g_2 \rangle</M> hold then we have
<M>d(s_1, g_1^G) = d(s_2, g_1^G) = d(s_1, g_2^G) = d(s_2, g_2^G)</M>,
so only one of these values must be computed.
</Item>
<Item>
If both <M>s</M> and <M>g</M> are contained in one of the normal subgroups given
then <M>d(s, g^G)</M> is zero.
</Item>
<Item>
If <M>G</M> is not a dihedral group and both <M>s</M> and <M>g</M> are involutions
then <M>d(s, g^G)</M> is zero.
</Item>
</List>
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F VertexDegreesGeneratingGraph( <G>, <classes>, <normalsubgroups> )
##
## Let <G> be a finite group that acts *transitively* on its moved points,
## <classes> be the list of conjugacy classes of <G>
## (in order to prescribe an ordering of the classes),
## and <normalsubgroups> be a list of proper normal subgroups of <G>.
##
## 'VertexDegreesGeneratingGraph' returns the matrix $d(g_i, g_j^G)$,
## with rows and columns indexed by nonidentity class representatives
## ordered as in the list <classes>.
## (The class containing the identity element may be contained in
## 'classes'.)
##
]]></Ignore>
<Example><![CDATA[
gap> BindGlobal( "VertexDegreesGeneratingGraph",
> function( G, classes, normalsubgroups )
> local nccl, matrix, cents, powers, normalsubgroupspos, i, j, g_i,
> nsg, g_j, gen, pair, d, pow;
>
> if not IsTransitive( G ) then
> Error( " must be transitive on its moved points" );
> fi;
>
> classes:= Filtered( classes,
> C -> Order( Representative( C ) ) <> 1 );
> nccl:= Length( classes );
> matrix:= [];
> cents:= [];
> powers:= [];
> normalsubgroupspos:= [];
> for i in [ 1 .. nccl ] do
> matrix[i]:= [];
> if IsBound( powers[i] ) then
> # The i-th row equals the earlier row 'powers[i]'.
> for j in [ 1 .. i ] do
> matrix[i][j]:= matrix[ powers[i] ][j];
> matrix[j][i]:= matrix[j][ powers[i] ];
> od;
> else
> # We have to compute the values.
> g_i:= Representative( classes[i] );
> nsg:= Filtered( [ 1 .. Length( normalsubgroups ) ],
> i -> g_i in normalsubgroups[i] );
> normalsubgroupspos[i]:= nsg;
> cents[i]:= Centralizer( G, g_i );
> for j in [ 1 .. i ] do
> g_j:= Representative( classes[j] );
> if IsBound( powers[j] ) then
> matrix[i][j]:= matrix[i][ powers[j] ];
> matrix[j][i]:= matrix[ powers[j] ][i];
> elif not IsEmpty( Intersection( nsg, normalsubgroupspos[j] ) )
> or ( Order( g_i ) = 2 and Order( g_j ) = 2
> and not IsDihedralGroup( G ) ) then
> matrix[i][j]:= 0;
> matrix[j][i]:= 0;
> else
> # Compute $d(g_i, g_j^G)$.
> gen:= 0;
> for pair in DoubleCosetRepsAndSizes( G, cents[j],
> cents[i] ) do
> if IsGeneratorsOfTransPermGroup( G,
> [ g_i, g_j^pair[1] ] ) then
> gen:= gen + pair[2];
> fi;
> od;
> matrix[i][j]:= gen / Size( cents[j] );
> if i <> j then
> matrix[j][i]:= gen / Size( cents[i] );
> fi;
> fi;
> od;
>
> # For later, provide information about algebraic conjugacy.
> for d in Difference( PrimeResidues( Order( g_i ) ), [ 1 ] ) do
> pow:= g_i^d;
> for j in [ i+1 .. nccl ] do
> if not IsBound( powers[j] ) and pow in classes[j] then
> powers[j]:= i;
> break;
> fi;
> od;
> od;
> fi;
> od;
>
> return matrix;
> end );
]]></Example>
In this section,
the task is to compute the lower bounds <M>\delta(s, g^G)</M> for the
vertex degrees <M>d(s, g^G)</M> using character-theoretic methods.
<P/>
We provide &GAP; functions for computing
the multiset <M>\Pi</M> of the primitive permutation characters
of a given group <M>G</M> and
for computing the lower bounds <M>\delta(s, g^G)</M> from <M>\Pi</M>.
<P/>
For many almost simple groups, the &GAP; libraries of character tables
and of tables of marks contain information for quickly computing
the primitive permutation characters of the group in question.
Therefore, the function <C>PrimitivePermutationCharacters</C> takes as its
argument not the group <M>G</M> but its character table <M>T</M>, say.
(This function is shown already in <Cite Key="ProbGenArxiv"/>.)
<P/>
If <M>T</M> is contained in the &GAP; Character Table Library
(see <Cite Key="CTblLib"/>)
then the complete set of primitive permutation characters
can be easily computed if the character tables of all maximal
subgroups and their class fusions into <M>T</M> are known
(in this case, we check whether the attribute
<Ref Func="Maxes" BookName="ctbllib"/>
of <M>T</M> is bound)
or if the table of marks of <M>G</M> and the class fusion from <M>T</M> into this
table of marks are known
(in this case, we check whether the attribute
<Ref Func="FusionToTom" BookName="ctbllib"/>
of <M>T</M> is bound).
If the attribute
<Ref Func="UnderlyingGroup" BookName="ref"/>
of <M>T</M> is bound then the group
stored as the value of this attribute
can be used to compute the primitive permutation characters.
The latter happens if <M>T</M> was computed from the group <M>G</M>;
for tables in the &GAP; Character Table Library,
this is not the case by default.
<P/>
The &GAP; function <C>PrimitivePermutationCharacters</C> tries to compute
the primitive permutation characters of a group using this information;
it returns the required list of characters if this can be computed this way,
otherwise <K>fail</K> is returned.
(For convenience, we use the &GAP; mechanism of <E>attributes</E>
in order to store the permutation characters in the character table object
once they have been computed.)
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#A PrimitivePermutationCharacters( <tbl> )
##
## For an ordinary character table <tbl> for which either the value of one
## of the attributes 'Maxes' or 'UnderlyingGroup' is stored or the table of
## marks is contained in the GAP library of tables of marks,
## 'PrimitivePermutationCharacters' returns the list of all primitive
## permutation characters of <tbl>.
## Otherwise 'fail' is returned.
##
## We use 'InstallOtherMethod' not 'InstallMethod' because another test file
## declares the same attribute and installs the same method.
##
]]></Ignore>
<Example><![CDATA[
gap> DeclareAttribute( "PrimitivePermutationCharacters",
> IsCharacterTable );
gap> InstallOtherMethod( PrimitivePermutationCharacters,
> [ IsCharacterTable ],
> function( tbl )
> local maxes, i, fus, poss, tom, G;
>
> if HasMaxes( tbl ) then
> maxes:= List( Maxes( tbl ), CharacterTable );
> for i in [ 1 .. Length( maxes ) ] do
> fus:= GetFusionMap( maxes[i], tbl );
> if fus = fail then
> fus:= PossibleClassFusions( maxes[i], tbl );
> poss:= Set( fus,
> map -> InducedClassFunctionsByFusionMap(
> maxes[i], tbl,
> [ TrivialCharacter( maxes[i] ) ], map )[1] );
> if Length( poss ) = 1 then
> maxes[i]:= poss[1];
> else
> return fail;
> fi;
> else
> maxes[i]:= TrivialCharacter( maxes[i] )^tbl;
> fi;
> od;
> return maxes;
> elif HasFusionToTom( tbl ) then
> tom:= TableOfMarks( tbl );
> maxes:= MaximalSubgroupsTom( tom );
> return PermCharsTom( tbl, tom ){ maxes[1] };
> elif HasUnderlyingGroup( tbl ) then
> G:= UnderlyingGroup( tbl );
> return List( MaximalSubgroupClassReps( G ),
> M -> TrivialCharacter( M )^tbl );
> fi;
>
> return fail;
> end );
]]></Example> <!-- % improve: if HasMaxes but not all fusions available? (Babymonster!) -->
<P/>
The next function computes the lower bounds <M>\delta(s, g^G)</M> from
the two lists <C>classlengths</C> of conjugacy class lengths of the group <M>G</M>
and <C>prim</C> of all primitive permutation characters of <M>G</M>.
(The first entry in <C>classlengths</C> is assumed to represent the class
containing the identity element of <M>G</M>.)
The return value is the matrix that contains in row <M>i</M> and column <M>j</M>
the value <M>\delta(s, g^G)</M>, where <M>s</M> and <M>g</M> are in the conjugacy classes
represented by the <M>(i+1)</M>-st and <M>(j+1)</M>-st column of <C>tbl</C>, respectively.
So the row sums of this matrix are the values <M>\delta(s)</M>.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F LowerBoundsVertexDegrees( <classlengths>, <prim> )
##
## For two lists <classlengths> of the conjugacy class lengths and <prim> of
## all primitive permutation characters of a group $G$, say,
## 'LowerBoundsVertexDegrees' returns a matrix <delta> such that
## '[i][j] = '$\delta(s, g^G)$ holds,
## for $s$ and $g$ in the 'i+1'-st and 'j+1'-st class of $G$,
## respectively.
##
## So the row sums in <delta> are the values $\delta(s)$.
##
]]></Ignore>
<Example><![CDATA[
gap> LowerBoundsVertexDegrees:= function( classlengths, prim )
> local sizes, nccl;
>
> nccl:= Length( classlengths );
> return List( [ 2 .. nccl ],
> i -> List( [ 2 .. nccl ],
> j -> Maximum( 0, classlengths[j] - Sum( prim,
> pi -> classlengths[j] * pi[j] * pi[i]
> / pi[1] ) ) ) );
> end;;
]]></Example>
</Subsection>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="clos">
<Heading>Evaluating the (Lower Bounds for the) Vertex Degrees</Heading>
In this section,
the task is to compute (lower bounds for) the vertex degrees of iterated
closures of a generating graph from (lower bounds for) the vertex degrees
of the graph itself,
and then to check the criteria of Pósa and Chvátal.
<P/>
The arguments of all functions defined in this section are
the list <C>classlengths</C> of conjugacy class lengths for the group <M>G</M>
(including the class of the identity element, in the first position)
and a matrix <C>bounds</C> of the values <M>d^{(i)}(s, g^G)</M>
or <M>\delta^{(i)}(s, g^G)</M>,
with rows and columns indexed by nonidentity class representatives
<M>s</M> and <M>g</M>, respectively.
Such a matrix is returned by the functions <C>VertexDegreesGeneratingGraph</C> or
<C>LowerBoundsVertexDegrees</C>, respectively.
<P/>
The function <C>LowerBoundsVertexDegreesOfClosure</C> returns the corresponding
matrix of the values <M>d^{(i+1)}(s, g^G)</M> or <M>\delta^{(i+1)}(s, g^G)</M>,
respectively.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F LowerBoundsVertexDegreesOfClosure( <classlengths>, <bounds> )
##
## Given the list <classlengths> of conjugacy class lengths for the group
## $G$ and a matrix <bounds> of the values $d^{(i)}(s, g^G)$ or
## $\delta^{(i)}(s, g^G)$,
## as is returned by 'VertexDegreesGeneratingGraph' or
## 'LowerBoundsVertexDegrees',
## the following function returns the corresponding matrix of the values
## $d^{(i+1)}(s, g^G)$ or $\delta^{(i+1)}(s, g^G)$, respectively.
##
]]></Ignore>
<Example><![CDATA[
gap> LowerBoundsVertexDegreesOfClosure:= function( classlengths, bounds )
> local delta, newbounds, size, i, j;
>
> delta:= List( bounds, Sum );
> newbounds:= List( bounds, ShallowCopy );
> size:= Sum( classlengths );
> for i in [ 1 .. Length( bounds ) ] do
> for j in [ 1 .. Length( bounds ) ] do
> if delta[i] + delta[j] >= size - 1 then
> newbounds[i][j]:= classlengths[ j+1 ];
> fi;
> od;
> od;
>
> return newbounds;
> end;;
]]></Example>
<P/>
Once the values <M>d^{(i)}(s, g^G)</M> or <M>\delta^{(i)}(s, g^G)</M> are known,
we can check whether Pósa's or Chvátal's criterion is satisfied
for the graph <M>&cl;^{(i)}(\Gamma(G))</M>,
using the function <C>CheckCriteriaOfPosaAndChvatal</C> shown below.
(Of course a <E>negative</E> result is meaningless in the case that only
lower bounds for the vertex degrees are used.)
<P/>
The idea is to compute the row sums of the given matrix,
and to compute the intervals <M>\{ L_k, L_k + 1, \ldots, U_k \}</M>
and <M>\{ L^{\prime}_k, L^{\prime}_k + 1, \ldots, U^{\prime}_k \}</M>
that were introduced in Section <Ref Subsect="critcheck"/>.
<P/>
The function <C>CheckCriteriaOfPosaAndChvatal</C> returns,
given the list of class lengths of <M>G</M> and the matrix of (bounds for the)
vertex degrees, a record with the components
<C>badForPosa</C> (the list of those pairs <M>[ L_k, U_k ]</M> with the property
<M>L_k \leq U_k</M>),
<C>badForChvatal</C> (the list of pairs of lower and upper bounds of
nonempty intervals where Chvátal's criterion may be violated),
and <C>data</C> (the sorted list of triples <M>[ \delta(g_k), |g_k^G|, \iota(k) ]</M>,
where <M>\iota(k)</M> is the row and column position of <M>g_k</M> in the matrix
<C>bounds</C>).
The ordering of class lengths must of course be compatible with the
ordering of rows and columns of the matrix,
and the identity element of <M>G</M> must belong to the first entry in the
list of class lengths.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F CheckCriteriaOfPosaAndChvatal( <classlengths>, <bounds> )
##
## Let <classlengths> be list of conjugacy class lengths of a group $G$,
## say, and <bounds> be a matrix of the values $d^{(i)}(s, g^G)$ or
## $\delta^{(i)}(s, g^G)$,
## as is returned for example by 'LowerBoundsVertexDegrees' or
## 'LowerBoundsVertexDegreesOfClosure'.
##
## 'CheckCriteriaOfPosaAndChvatal' returns a record with the following
## components.
##
## 'badForPosa':
## the list of those pairs $[ L_k, U_k ]$ with the property
## $L_k \leq U_k$,
##
## 'badForChvatal':
## the list of pairs of lower and upper bounds of nonempty intervals
## where Chvátal's criterion may be violated, and
##
## 'data':
## the sorted list of triples $[ \delta(g_k), |g_k^G|, \iota(k) ]$,
## where $\iota(k)$ is the row and column position of $g_k$
## in the matrix <bounds>.
##
## The generating graph $\Gamma(G)$ satisfies Pósa's criterion
## if the 'badForPosa' component is empty;
## the graph satisfies Chvátal's criterion if the 'badForChvatal'
## component is empty.
##
## The ordering of class lengths must of course be compatible with the
## ordering of rows and columns of the matrix,
## and the identity element of $G$ must belong to the first entry in the
## list of class lengths.
##
]]></Ignore>
<Example><![CDATA[
gap> CheckCriteriaOfPosaAndChvatal:= function( classlengths, bounds )
> local size, degs, addinterval, badForPosa, badForChvatal1, pos,
> half, i, low1, upp2, upp1, low2, badForChvatal, interval1,
> interval2;
>
> size:= Sum( classlengths );
> degs:= List( [ 2 .. Length( classlengths ) ],
> i -> [ Sum( bounds[ i-1 ] ), classlengths[i], i ] );
> Sort( degs );
>
> addinterval:= function( intervals, low, upp )
> if low <= upp then
> Add( intervals, [ low, upp ] );
> fi;
> end;
>
> badForPosa:= [];
> badForChvatal1:= [];
> pos:= 1;
> half:= Int( size / 2 ) - 1;
> for i in [ 1 .. Length( degs ) ] do
> # We have pos = c_1 + c_2 + \cdots + c_{i-1} + 1
> low1:= Maximum( pos, degs[i][1] ); # L_i
> upp2:= Minimum( half, size-1-pos, size-1-degs[i][1] ); # U'_i
> pos:= pos + degs[i][2];
> upp1:= Minimum( half, pos-1 ); # U_i
> low2:= Maximum( 1, size-pos ); # L'_i
> addinterval( badForPosa, low1, upp1 );
> addinterval( badForChvatal1, low2, upp2 );
> od;
>
> # Intersect intervals.
> badForChvatal:= [];
> for interval1 in badForPosa do
> for interval2 in badForChvatal1 do
> addinterval( badForChvatal,
> Maximum( interval1[1], interval2[1] ),
> Minimum( interval1[2], interval2[2] ) );
> od;
> od;
>
> return rec( badForPosa:= badForPosa,
> badForChvatal:= Set( badForChvatal ),
> data:= degs );
> end;;
]]></Example>
<P/>
Finally, the function <C>HamiltonianCycleInfo</C> assumes that the matrix <C>bounds</C>
contains lower bounds for the vertex degrees in the generating graph
<M>\Gamma</M>, and returns a string that describes the minimal <M>i</M> with the
property that the given bounds suffice to show that <M>cl^{(i)}(\Gamma)</M>
satisfies Pósa's or Chvátal's criterion,
if such a closure exists.
If no closure has this property, the string <C>"no decision"</C> is returned.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F HamiltonianCycleInfo( <classlengths>, <bounds> )
##
## Let <classlengths> be list of conjugacy class lengths of a group $G$,
## say, and <bounds> be a matrix of the values $d^{(i)}(s, g^G)$ or
## $\delta^{(i)}(s, g^G)$,
## as is returned for example by 'LowerBoundsVertexDegrees' or
## 'LowerBoundsVertexDegreesOfClosure'.
##
## 'HamiltonianCycleInfo' returns a string that describes the minimal $i$
## with the property that the given bounds suffice to show that
## $cl^{(i)}(\Gamma(G))$ satisfies Pósa's or Chvátal's criterion,
## if such a closure exists.
## If no closure has this property, the string '"no decision"' is returned.
##
]]></Ignore>
<Example><![CDATA[
gap> HamiltonianCycleInfo:= function( classlengths, bounds )
> local i, result, res, oldbounds;
>
> i:= 0;
> result:= rec( Posa:= fail, Chvatal:= fail );
> repeat
> res:= CheckCriteriaOfPosaAndChvatal( classlengths, bounds );
> if result.Posa = fail and IsEmpty( res.badForPosa ) then
> result.Posa:= i;
> fi;
> if result.Chvatal = fail and IsEmpty( res.badForChvatal ) then
> result.Chvatal:= i;
> fi;
> i:= i+1;
> oldbounds:= bounds;
> bounds:= LowerBoundsVertexDegreesOfClosure( classlengths,
> bounds );
> until oldbounds = bounds;
>
> if result.Posa <> fail then
> if result.Posa <> result.Chvatal then
> return Concatenation(
> "Chvatal for ", Ordinal( result.Chvatal ), " closure, ",
> "Posa for ", Ordinal( result.Posa ), " closure" );
> else
> return Concatenation( "Posa for ", Ordinal( result.Posa ),
> " closure" );
> fi;
> elif result.Chvatal <> fail then
> return Concatenation( "Chvatal for ", Ordinal( result.Chvatal ),
> " closure" );
> else
> return "no decision";
> fi;
> end;;
]]></Example>
In this section, we apply the functions introduced in Section <Ref Subsect="functions"/>
to character tables of almost simple groups
that are available in the &GAP; Character Table Library.
<P/>
Our first examples are the sporadic simple groups, in Section <Ref Subsect="spor"/>,
then their automorphism groups are considered in Section <Ref Subsect="sporaut"/>.
Small alternating and symmetric groups are treated in
Section <Ref Subsect="symmalt"/>.
<P/>
For our convenience, we provide a small function that takes as its
argument only the character table in question,
and returns a string, either <C>"no prim. perm. characters"</C> or the
return value of <C>HamiltonianCycleInfo</C> for the bounds computed from
the primitive permutation characters.
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F HamiltonianCycleInfoFromCharacterTable( <tbl> )
##
## For a character table <tbl>, this function returns a string,
## either '"no prim. perm. characters"' or the return value of
## 'HamiltonianCycleInfo' for the bounds computed from the primitive
## permutation characters.
##
]]></Ignore>
<Example><![CDATA[
gap> HamiltonianCycleInfoFromCharacterTable:= function( tbl )
> local prim, classlengths, bounds;
>
> prim:= PrimitivePermutationCharacters( tbl );
> if prim = fail then
> return "no prim. perm. characters";
> fi;
> classlengths:= SizesConjugacyClasses( tbl );
> bounds:= LowerBoundsVertexDegrees( classlengths, prim );
> return HamiltonianCycleInfo( classlengths, bounds );
> end;;
]]></Example>
Since version 1.3.10 of &CTblLib;,
the &GAP; Character Table Library contains the tables of maximal subgroups
of all sporadic simple groups.
<P/>
<!-- % For the Baby Monster, --> <!-- % currently the class fusion of the 6-th maximal subgroup is not known. --> <!-- % It takes several hours of CPU time to verify that the permutation --> <!-- % character is uniquely determined by the character tables. -->
So the function <C>PrimitivePermutationCharacters</C>
can be used to compute all their primitive permutation characters.
It turns out that the information available in the
&GAP; Character Table Library is sufficient to prove that
the generating graph contains a Hamiltonian cycle.
In the original version of this file, a special treatment of the Monster
group had been necessary because not all character tabes of its maximal
subgroups were available, and also not all class fusions from some
maximal subgroups were known whose character tables were available.
This section shows the computations that were used at that time.
<P/>
Meanwhile the maximal subgroups of the Monster are classified,
see <Cite Key="DLP25"/>, their character tables are available,
and also all class fusions from the maximal subgroups are known.
Thus we can easily compute the primitive permutation characters
of the Monster group.
The computations shown in this section are not needed to prove the claims
from <Cite Key="GMN"/>, they are kept just for historical reasons.
Thus this section is an example how the availability of the full list of
primitive permutation characters of the Monster simplifies proofs.
<P/>
The situation for the following arguments is as follows.
We do not assume the knowledge of all primitive permutation characters of the
Monster group <M>M</M>, and we use only some of the character tables of the
maximal subgroups of <M>M</M> and only partial information on their
class fusions.
Note that for some maximal subgroups of <M>M</M>,
the permutation character is not uniquely determined by the character tables
involved.
However, we can compute upper bounds for the values of the primitive
permutation characters <M>1_S^M</M> from the possible class fusions
from <M>S</M> into <M>M</M> if the character table of <M>S</M> is known.
For the other subgroups <M>S</M>, the permutation characters <M>1_S^M</M>
have been computed with other methods.
Using this information, we will show that the generating graph of <M>M</M>
satisfies Pósa's criterion.
<P/>
The list <C>primdata</C> defined below has length <M>46</M>.
The entry at position <M>i</M> is a list of length one or two.
If <C>primdata[</C><M>i</M><C>]</C> has length one then its unique entry is
the identifier of the library character table of the <M>i</M>-th maximal
subgroup of <M>M</M>.
If <C>primdata[</C><M>i</M><C>]</C> has length two then its entries are
a string describing the structure of the <M>i</M>-th maximal subgroup
<M>S</M> of <M>M</M> and the permutation character <M>1_S^M</M>.
<P/>
(The construction of the explicitly given characters in this list
will be documented elsewhere.
Some of the constructions can be found
in Section <Ref Sect="sect:monsterperm"/>.)
We compute upper bounds for the permutation character values
in the cases where the characters are not given explicitly.
(We could improve this by using additional information about the class
fusions, but this will not be necessary.)
<P/>
<Example><![CDATA[
gap> s:= "dummy";; # Avoid a message about an unbound variable ...
gap> poss:= "dummy";; # Avoid a message about an unbound variable ...
gap> for entry in primdata do
> if not IsBound( entry[2] ) then
> s:= CharacterTable( entry[1] );
> poss:= Set( PossibleClassFusions( s, m ),
> x -> InducedClassFunctionsByFusionMap( s, m,
> [ TrivialCharacter( s ) ], x )[1] );
> entry[2]:= List( [ 1 .. NrConjugacyClasses( m ) ],
> i -> Maximum( List( poss, x -> x[i] ) ) );
> fi;
> od;
]]></Example>
<!-- This loop required about half an hour of CPU time in 2012,
but just six minutes in 2023. -->
<P/>
Now we estimate the lower bounds <M>\delta(s, g^G)</M> introduced in
Section <Ref Subsect="characters"/>.
Let <M>&M;</M> denote a set of representatives of the
classes of maximal subgroups of <M>M</M>.
Then
hence <M>\delta(s)</M> can be computed from the corresponding primitive
permutation characters,
and a lower bound for <M>\delta(s)</M> can be computed from the upper bounds for
the characters <M>1_S^G</M> which are given by the list <C>primdata</C>.
<P/>
This means that modifying the output of <C>LowerBoundsVertexDegrees</C>
as follows really yields lower bounds for the vertex degrees.
<P/>
<Example><![CDATA[
gap> prim:= List( primdata, x -> x[2] );;
gap> classlengths:= SizesConjugacyClasses( m );;
gap> bounds:= LowerBoundsVertexDegrees( classlengths, prim );;
]]></Example>
<P/>
Now we sum up the bounds for the individual classes.
It turns out that the minimal vertex degree is more than <M>99.99998</M>
percent of <M>|M|</M>.
This proves that the generating graph of the Monster
satisfies Pósa's criterion.
<E>Without</E> the results from <Cite Key="DLP25"/>,
we can argue as follows.
(This was the situation in earlier versions of this example file.)
<P/>
According to <Cite Key="NW12"/>,
any maximal subgroup of the Monster is either among the <M>44</M> known
classes from the above list except L<M>_2(13).2</M> and U<M>_3(4).4</M>,
or it is an almost simple group whose socle is one of
L<M>_2(13)</M>, Sz<M>(8)</M>, U<M>_3(4)</M>, and U<M>_3(8)</M>. <!-- % -> has been ruled out: L<M>_2(27)</M>, -->
<P/>
We show that the elements of such subgroups are contained in
the union of <M>55</M> conjugacy classes of the Monster
that cover less than one percent of the elements in the Monster.
For that, we compute the possible class fusions from the abovementioned
simple groups <M>S</M> into the Monster, and then the possible class fusions
from the automorphic extensions of <M>S</M> into the Monster,
using the possible class fusions of <M>S</M>.
(This approach is faster than computing each class fusion from scratch.)
<P/>
After the following computations,
the list <C>badclasses</C> will contain the positions of all those classes
of <M>M</M> that may contain elements in some of the hypothetical
maximal subgroups.
<P/>
For each simple group in question, we enter the identifiers of the
character tables of the automorphic extensions that can occur.
Note that the automorphism groups of the four groups have the structures
L<M>_2(13).2</M>, Sz<M>(8).3</M>, U<M>_3(4).4</M>,
and U<M>_3(8).(3 \times S_3)</M>, respectively.
We need not consider the groups U<M>_3(8).3^2</M> and
U<M>_3(8).(3 \times S_3)</M>
because already U<M>_3(8).3_2</M> does not admit an embedding into <M>M</M>,
and we need not consider the group U<M>_3(8).S_3</M> because its set of
elements is covered by its subgroups of the types U<M>_3(8).2</M> and
U<M>_3(8).3_2</M>.
<E>In the original version of this file, also hypothetical maximal
subgroups with socle</E> L<M>_2(27)</M> <E>had been considered.
As a consequence, the list</E> <C>badclasses</C> <E>computed above had length
<M>59</M> in the original version;
the list contained also the classes at the positions <M>90, 94, 95</M>,
and <M>96</M>,
that is, the classes</E> <C>26B</C>, <C>28B</C>, <C>28C</C>, <C>28D</C>.
<E>The proportion</E> <C>bad</C> <E>of elements in the classes of <M>M</M>
described by</E>
<C>badclasses</C> <E>was about <M>2.05</M> percent of <M>|M|</M>,
compared to the about <M>0.98</M> percent in the current version.</E>
<!-- % If we ignore L<M>_2(27)</M> then we do not get the classes [ 90, 94, 95, 96 ], --> <!-- % i.e., [ "26b", "28b", "28c", "28d" ]. -->
<P/>
Now we estimate the lower bounds <M>\delta(s, g^G)</M> introduced in
Section <Ref Subsect="characters"/>.
Let <M>&B;</M> denote the union of the classes described by <C>badclasses</C>,
and let <M>&M;</M> denote a set of representatives of the
<M>44</M> known classes of maximal subgroups of <M>M</M>.
hence <M>\delta(s)</M> can be computed from the corresponding primitive
permutation characters,
and a lower bound for <M>\delta(s)</M> can be computed from the upper bounds for
the characters <M>1_S^G</M> which are given by the list <C>primdata</C>.
<P/>
If <M>s \in &B;</M> then the above equation for <M>\delta(s, g^G)</M> holds
at least for <M>g \notin &B;</M>,
so <M>\sum_{{g \in R \setminus &B;}} \delta(s, g^G)</M> is a lower bound
for <M>\delta(s)</M>.
So <C>primdata</C> yields a lower bound for <M>\delta(s)</M> also for
<M>s \in &B;</M>,
by ignoring the pairs <M>(s, g)</M> where both <M>s</M> and <M>g</M> lie in
<M>&B;</M>.
<P/>
This means that modifying the output of <C>LowerBoundsVertexDegrees</C>
as follows really yields lower bounds for the vertex degrees.
(Note that the row and column positions in the matrix returned by
<C>LowerBoundsVertexDegrees</C> are shifted by one,
compared to <C>badclasses</C>.)
<P/>
<Example><![CDATA[
gap> prim:= List( primdata, x -> x[2] );;
gap> badpos:= Difference( badclasses, [ 1 ] ) - 1;;
gap> bounds:= LowerBoundsVertexDegrees( classlengths, prim );;
gap> for i in badpos do
> for j in badpos do
> bounds[i][j]:= 0;
> od;
> od;
]]></Example>
<P/>
Now we sum up the bounds for the individual classes.
It turns out that the minimal vertex degree is more than <M>99</M> percent
of <M>|M|</M>.
This proves that the generating graph of the Monster
satisfies Pósa's criterion.
<P/>
(And the minimal vertex degree of elements outside <M>&B;</M>
is more than <M>99.99998</M> percent of <M>|M|</M>.)
<P/>
<E>In the original version of this file,
we got only <M>97.95</M> percent of <M>|M|</M> as the lower bound
for the minimal vertex degree.
The bound for elements outside <M>&B;</M> was the same
in the original version.
The fact that the maximal subgroups of type</E> L<M>_2(41)</M>
<E>had been ignored in the original version
did not affect the lower bound for the minimal vertex degree.</E>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="sporaut">
<Heading>Nonsimple Automorphism Groups of Sporadic Simple Groups</Heading>
Next we consider the nonsimple automorphism groups of the sporadic simple
groups.
Nontrivial outer automorphisms exist exactly in <M>12</M> cases,
and then the simple group has index <M>2</M> in its automorphism group.
The character tables of the groups and their maximal subgroups are
available in &GAP;.
<P/>
<Example><![CDATA[
gap> spornames:= AllCharacterTableNames( IsSporadicSimple, true,
> IsDuplicateTable, false );;
gap> sporautnames:= AllCharacterTableNames( IsSporadicSimple, true,
> IsDuplicateTable, false,
> OfThose, AutomorphismGroup );;
gap> sporautnames:= Difference( sporautnames, spornames );
[ "F3+.2", "Fi22.2", "HN.2", "HS.2", "He.2", "J2.2", "J3.2", "M12.2", "M22.2", "McL.2", "ON.2", "Suz.2" ]
gap> for tbl in List( sporautnames, CharacterTable ) do
> info:= HamiltonianCycleInfoFromCharacterTable( tbl );
> Print( Identifier( tbl ), ": ", info, "\n" );
> od;
F3+.2: Chvatal for 0th closure, Posa for 1st closure
Fi22.2: Chvatal for 0th closure, Posa for 1st closure
HN.2: Chvatal for 0th closure, Posa for 1st closure
HS.2: Chvatal for 1st closure, Posa for 2nd closure
He.2: Chvatal for 0th closure, Posa for 1st closure
J2.2: Chvatal for 0th closure, Posa for 1st closure
J3.2: Chvatal for 0th closure, Posa for 1st closure
M12.2: Chvatal for 0th closure, Posa for 1st closure
M22.2: Posa for 1st closure
McL.2: Chvatal for 0th closure, Posa for 1st closure
ON.2: Chvatal for 0th closure, Posa for 1st closure
Suz.2: Chvatal for 0th closure, Posa for 1st closure
]]></Example>
</Subsection>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="symmalt">
<Heading>Alternating and Symmetric Groups <M>A_n</M>, <M>S_n</M>,
for <M>5 \leq n \leq 13</M></Heading>
For alternating and symmetric groups <M>A_n</M> and <M>S_n</M>, respectively,
with <M>5 \leq n \leq 13</M>,
the table of marks or the character tables of the group
and all its maximal subgroups are available in &GAP;.
So we can compute the character-theoretic bounds for vertex degrees.
<P/>
<Example><![CDATA[
gap> for tbl in List( [ 5 .. 13 ], i -> CharacterTable(
> Concatenation( "A", String( i ) ) ) ) do
> info:= HamiltonianCycleInfoFromCharacterTable( tbl );
> if info <> "Posa for 0th closure" then
> Print( Identifier( tbl ), ": ", info, "\n" );
> fi;
> od;
]]></Example>
<P/>
No messages are printed, so the generating graphs of the alternating
groups in question satisfy Pósa's criterion.
<P/>
<Example><![CDATA[
gap> for tbl in List( [ 5 .. 13 ], i -> CharacterTable(
> Concatenation( "S", String( i ) ) ) ) do
> info:= HamiltonianCycleInfoFromCharacterTable( tbl );
> Print( Identifier( tbl ), ": ", info, "\n" );
> od;
A5.2: no decision
A6.2_1: Chvatal for 4th closure, Posa for 5th closure
A7.2: Posa for 1st closure
A8.2: Chvatal for 2nd closure, Posa for 3rd closure
A9.2: Chvatal for 2nd closure, Posa for 3rd closure
A10.2: Chvatal for 2nd closure, Posa for 3rd closure
A11.2: Posa for 1st closure
A12.2: Chvatal for 2nd closure, Posa for 3rd closure
A13.2: Posa for 1st closure
]]></Example>
<P/>
We see that sufficiently large closures of the generating graphs of the
symmetric groups in question satisfy Pósa's criterion,
except that the bounds for the symmetric group <M>S_5</M> are not sufficient
for the proof.
In Section <Ref Subsect="smallalmostsimple"/>,
it is shown that the 2nd closure of the generating graph of <M>S_5</M>
satisfies Pósa's criterion.
<P/>
(We could find slightly better bounds derived only from character tables
which suffice to prove that the generating graph for <M>S_5</M> contains a
Hamiltonian cycle, but this seems to be not worth while.)
</Subsection>
</Section>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Section Label="grouptheor">
<Heading>Computations With Groups</Heading>
We prove in Section <Ref Subsect="smallsimp"/> that the generating
graphs of the nonabelian simple groups of order up to <M>10^6</M>
satisfy Pósa's criterion,
and that the same holds for those nonabelian simple
groups of order between <M>10^6</M> and <M>10^7</M> that are not isomorphic
with some <M>&PSL;(2,q)</M>.
(In Section <Ref Subsect="psl2q"/>,
it is shown that the generating graph of
<M>&PSL;(2,q)</M> satifies Pósa's criterion for any prime power q.)
Nonsimple nonsolvable groups of order up to <M>10^6</M>
are treated in Section <Ref Subsect="smallalmostsimple"/>.
<P/>
(We could increase the bounds <M>10^6</M> and <M>10^7</M> with more
computations, using the same methods.)
<P/>
For our convenience, we provide a small function that takes as its
argument only the group in question,
and returns a string, the
return value of <C>HamiltonianCycleInfo</C> for the vertex degrees computed
from the group.
(In order to speed up the computations,
the function computes the proper normal subgroups that contain the derived
subgroup of the given group, and enters the list of these groups as the
third argument of <C>VertexDegreesGeneratingGraph</C>.)
<P/>
<Ignore Remark="gapfilecomments"><![CDATA[
#############################################################################
##
#F HamiltonianCycleInfoFromGroup( <G> )
##
## For a group <G>, this function returns a string,
## the return value of 'HamiltonianCycleInfo' for the vertex degrees
## computed from the group.
##
]]></Ignore>
<Example><![CDATA[
gap> HamiltonianCycleInfoFromGroup:= function( G )
> local ccl, nsg, der, degrees, classlengths;
> ccl:= ConjugacyClasses( G );
> if IsPerfect( G ) then
> nsg:= [];
> else
> der:= DerivedSubgroup( G );
> nsg:= Concatenation( [ der ],
> IntermediateSubgroups( G, der ).subgroups );
> fi;
> degrees:= VertexDegreesGeneratingGraph( G, ccl, nsg );
> classlengths:= List( ccl, Size );
> return HamiltonianCycleInfo( classlengths, degrees );
> end;;
]]></Example>
<!-- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -->
<Subsection Label="smallsimp">
<Heading>Nonabelian Simple Groups of Order up to <M>10^7</M></Heading>
Representatives of the <M>56</M> isomorphism types of nonabelian simple groups
of order up to <M>10^6</M> can be accessed in &GAP; with the function
<C>AllSmallNonabelianSimpleGroups</C>.
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung ist noch experimentell.
