The number of immediate descendants of algebra 4.1 of order $p^{7}$ is 1361
if $p=3$. For $p>3$ it is $p^{5}+2p^{4}+7p^{3}+25p^{2}+88p+270+(p+4)\gcd
(p-1,3)+\gcd (p-1,4)$.
If $L$ is an immediate descendant of 4.1 of order $p^7$ then $L$ is
generated by $a,b,c,d$, $L_2$ has order $p^3$, and $L_3=\{0\}$.
If $L^{2}$ has order $p$ then we can assume that $L^{2}$ is generated by $ba$
and that one of the following two sets of commutator relations hold: \begin{eqnarray*}
ca &=&da=cb=db=dc=0, \\
ca &=&da=cb=db=0,\,dc=ba. \end{eqnarray*}%
There are 7 algebras in the first case, and 4 in the second case.
\section{$L^{2}$ has order $p^{2}$}
If $L^{2}$ has order $p^{2}$ then we can assume that one of the following
sets of commutator relations holds: \begin{eqnarray*}
da &=&cb=db=dc=0, \\
ca &=&da=cb=db=0, \\
da &=&cb=dc=0,\,db=ca, \\
da &=&cb=0,\,db=ca,\,dc=\omega ba. \end{eqnarray*}%
Note that $L^{2}$ is generated by $ba,ca$ in all but the second of these
algebras. In the second algebra, $L^{2}$ is generated by $ba,dc$. We obtain $%
2p+29$ algebras in the first case, $(p^{2}-1)/2+4p+30$ in the second, $3p+26$
in the third, and $(p^{2}-1)/2+2p+6$ in the fourth.
In solving the isomorphism problem in Case 4, we have the following
presentation:% \[ \langle a,b,c,d\,|\,da,cb,db-ca,dc-\omega ba,pa,pb-xba-yca,pc-zba-tca,\,% \text{class }2\rangle , \]%
where $\left( \begin{array}{ll}
x & y \\
z & t% \end{array}% \right) $ runs over a set of representatives for the equivalence classes of
non-singular matrices $A$ under the equivalence relation given by% \[
A\sim\alpha ^{-1}\left( \begin{array}{ll} \mu & \nu\\ \pm\omega\nu & \pm\mu% \end{array}% \right) A\left( \begin{array}{ll} \mu & \nu\\ \pm\omega\nu & \pm\mu% \end{array}% \right) ^{-1}. \]%
There are $(p+1)^{2}/2$ equivalence classes.
There is a \textsc{Magma} program in notes4.1case4.m to compute a set of
representative matrices $A$.
\section{$L^{2}$ has order $p^{3}$}
If $L^2$ has order $p^3$ then $L$ must have the same commutator structure as
one of 7.15 -- 7.20 from the list of nilpotent Lie algebras of dimension 7
over $\mathbb{Z}_p$, so we can assume that one of the following sets of
commutator relations holds: \begin{eqnarray*}
da &=&db=dc=0, \\
ca &=&da=db=0, \\
ca &=&da=dc=0, \\
ca &=&da=0,\,dc=ba, \\
da &=&0,\,db=ca,\,dc=cb, \\
da &=&0,\,db=\omega ca,\,dc=ba. \end{eqnarray*}
In Case 1 we have $3p+18$ algebras.
In Case 2 we have $\allowbreak\frac{77}{2}p+\frac{173}{2}+11p^{2}+\frac{5}{2%
}p^{3}+\frac{1}{2}p^{4}$ algebras, but you need to add 2 if $p=1\func{mod}3$.
In Case 3 we have $\allowbreak p^{2}+3p+15$, but again you need to add 2 if $%
p=1\func{mod}3$.
In Case 4 we have $3p^{2}+13p+31$ algebras, but we need to add 2 if $p=1% \func{mod}4$ and add 2 if $p=1\func{mod}3$.
In Case 5 we have 550 algebras when $p=3$ and \begin{eqnarray*} \allowbreak p^{5}+p^{4}+4p^{3}+6p^{2}+18p+19\text{ if }p &=&1\func{mod}3, \\ \allowbreak p^{5}+p^{4}+4p^{3}+6p^{2}+16p+17\text{ if }p &=&2\func{mod}3. \end{eqnarray*}
In Case 6 we have $\frac{9}{2}p+\frac{13}{2}+3p^{2}+\frac{1}{2}p^{4}+\frac{1%
}{2}p^{3}$ algebras.
We need computer programs to sort out the isomorphism problem in Case 5 and
in Case 6.
\subsection{Case 5}
Let $L$ satisfy $da=0,\,db=ca,\,dc=cb$. It is convenient to replace $b$ by $%
b+d$, so that $L$ satisfies $da=cb=0$, $db=ca$. So $L^{2}$ is generated by $%
ba$, $ca$ and $dc$, and $pL\leq L^{2}$. It is fairly easy to see that if $%
a^{\prime },b^{\prime },c^{\prime },d^{\prime }$ generate $L$ and satisfy $%
d^{\prime }a^{\prime }=c^{\prime }b^{\prime }=0$, $d^{\prime }b^{\prime
}=c^{\prime }a^{\prime }$, then (modulo $L^{2}$) \begin{eqnarray*}
a^{\prime } &=&\alpha\lambda a+\beta\lambda b+\beta\mu c-\alpha\mu d, \\
b^{\prime } &=&\gamma\lambda a+\delta\lambda b+\delta\mu c-\gamma\mu d, \\
c^{\prime } &=&\gamma\nu a+\delta\nu b+\delta\xi c-\gamma\xi d, \\
d^{\prime } &=&-\alpha\nu a-\beta\nu b-\beta\xi c+\alpha\xi d \end{eqnarray*}%
with $(\alpha ,\beta )$ and $(\gamma ,\delta )$ linearly independant, and
with $(\lambda ,\mu )$ and $(\nu ,\xi )$ linearly independant. Furthermore \[ \left( \begin{array}{c}
b^{\prime }a^{\prime } \\
c^{\prime }a^{\prime } \\
d^{\prime }c^{\prime }% \end{array}% \right) =(\alpha\delta -\beta\gamma )\left( \begin{array}{ccc} \lambda ^{2} & 2\lambda\mu & \mu ^{2} \\ \lambda\nu & \lambda\xi +\mu\nu & \mu\xi\\ \nu ^{2} & 2\nu\xi & \xi ^{2}% \end{array}% \right) \left( \begin{array}{c}
ba \\
ca \\
dc% \end{array}% \right) . \]%
So we consider orbits of $4\times 3$ matrices $A$ (representing $pa,pb,pc,pd$%
) under transformations of the form \[
A\longmapsto (\alpha\delta -\beta\gamma )^{-1}\left( \begin{array}{cccc} \alpha\lambda & \beta\lambda & \beta\mu & -\alpha\mu\\ \gamma\lambda & \delta\lambda & \delta\mu & -\gamma\mu\\ \gamma\nu & \delta\nu & \delta\xi & -\gamma\xi\\
-\alpha\nu & -\beta\nu & -\beta\xi & \alpha\xi% \end{array}% \right) A\left( \begin{array}{ccc} \lambda ^{2} & 2\lambda\mu & \mu ^{2} \\ \lambda\nu & \lambda\xi +\mu\nu & \mu\xi\\ \nu ^{2} & 2\nu\xi & \xi ^{2}% \end{array}% \right) ^{-1}. \]%
We note that if we multiply $\alpha ,\beta ,\gamma ,\delta $ through by a
factor $k$ (in the expression above), and multiply $\lambda ,\mu ,\nu ,\xi $
through by a factor $l$, then the image of $A$ is multiplied by a factor $%
k^{-1}l^{-1}$. So we can ignore the factor $(\alpha\delta -\beta\gamma
)^{-1}$ and still get the same orbits.
We actually have an action of GL$(2,p)\times\,$GL$(2,p)$ on the vector
space of $4\times 3$ matrices, and if we leave out the factor $(\alpha \delta -\beta\gamma )^{-1}$ (as described above) then the kernel of the
action is the subgroup $\{(kI,kI)\,|$\thinspace $k\neq 0\}$, so (in effect)
we have a group of order $p^{2}\left( p-1\right) ^{3}\left( p+1\right) ^{2}$
acting on a space of order $p^{12}$.
If we take $\mu =0$ in the matrices above, then we obtain a subgroup $H$ of
the automorphism group of index $p+1$. There are \[
f(p)=p^{6}+2p^{5}+4p^{4}+8p^{3}+15p^{2}+29p+27+(2p+3)\gcd (p-1,3) \]%
orbits of matrices under the action of $H$, and we can \textquotedblleft
write down\textquotedblright\ a set of representatives for these orbits.
However for $p=19$ this takes about 3 minutes on my 5 year old linux box,
and the representatives take up 4.5 gigabytes of space. So I save space by
not writing all the representatives down in the program to generate orbit
representatives under the action of the full group $G$.
There is a \textsc{Magma} program to compute a set of orbit representatives
under the action of the full group $G$ in notes4.1case5.m. The
representatives are stored as $4\times 3$ matrices over GF$(p)$, which takes
up less space than storing them as integer sequences. We compute a
tranversal for the subgroup $H$ in $G$, and for each of the $f(p)$ $H$-orbit
representatives $A$, we compute the images of $A$ under elements of the
transversal, and determine how the $H$-orbits fuse under the action of $G$%
.Thus we have to consider $(p+1)f(p)$ matrices $At$ where $A$ is an $H$%
-orbit representative and $t$ is an element of the transversal. For each
such matrix $At$ we compute the $H$-orbit representative of $At$. (This
takes a bounded amount of work involving arithmetic over GF$(p)$.) We index
the $H$-orbits, and we add an $H$-orbit representative $A$ to the list of
the $G$-orbit representatives if the index of the $H$-orbit containing $A$
is greater than or equal to the indexes of the $H$-orbits containing the
matrices $At$ for $t$ in the transversal. So, if the index of the $H$-orbit
containing $At$ is less than the index of the $H$-orbit containing $A$, then
we discard $A$ and there is no need to consider the elements $Au$ for $u$ in
the remainder of the transversal. This means that we don't actually have to
consider all the elements $At$. For $p=3$ we only need to consider less than
two thirds of the elements $At$, for $p=5$ less than a half, for $p=7$ a
little over a third, and so on. Experimentally, it seems that the proportion
drops as the prime increases. So the total amount of work needed to compute
a set of representatives for the $G$-orbits is of order somewhere between $%
p^{6}$ and $p^{7}$. For $p\leq 23$ the time taken for the program to run is
roughly proportional to $p^{6.2}$. However this is a serious bottleneck, and
it takes about two hours to generate the list for $p=19$ on my five year old
linux box. Note however that $19^{5}=\allowbreak 2476\,099$, and there is
probably only a limited amount of interesting work you can do with two and
half million groups of order $19^{7}$.
\subsection{Case 6}
Let $L$ satisfy $da=0,\,db=\omega ca,\,dc=ba$. Then $L^{2}$ is generated by $%
ba$, $ca$, $cb$ and $pL\leq L^{2}$. It is straightforward to show that all
elements in the linear span of $a,b,c,d$ have breadth 3, except for those of
the form $\alpha a+\delta d$. Using this we can show that if $a^{\prime
},b^{\prime },c^{\prime },d^{\prime }$ generate $L$ and satisfy the same
commutator relations as $a,b,c,d$ then (modulo $L^{2}$) \begin{eqnarray*}
a^{\prime } &=&\alpha a+\delta d, \\
b^{\prime } &=&\pm (\lambda a+\gamma b+\omega\beta c+\mu d), \\
c^{\prime } &=&\nu a+\beta b+\gamma c+\xi d, \\
d^{\prime } &=&\pm (\omega\delta a+\alpha d) \end{eqnarray*}%
and \[ \left( \begin{array}{c}
b^{\prime }a^{\prime } \\
c^{\prime }a^{\prime } \\
c^{\prime }b^{\prime }% \end{array}% \right) =\left( \begin{array}{ccc} \pm (\alpha\gamma -\omega\beta\delta ) & \pm (\omega\alpha\beta -\omega \gamma\delta ) & 0 \\ \alpha\beta -\gamma\delta & \alpha\gamma -\omega\beta\delta & 0 \\ \pm (\beta\lambda -\gamma\nu +\omega\beta\xi -\gamma\mu ) & \pm (\gamma \lambda -\omega\beta\mu +\omega\gamma\xi -\omega\beta\nu ) & \pm
(\gamma ^{2}-\omega\beta ^{2})% \end{array}% \right) \left( \begin{array}{c}
ba \\
ca \\
cb% \end{array}% \right) . \]%
We let \[ \left( \begin{array}{c}
pa \\
pb \\
pc \\
pd% \end{array}% \right) =A\left( \begin{array}{c}
ba \\
ca \\
cb% \end{array}% \right) \]%
where $A$ is a $4\times 3$ matrix over $\mathbb{Z}_{p}$. Then under a change
of generating set of the form described above we see that \[
A\mapsto\left( \begin{array}{cccc} \alpha & 0 & 0 & \delta\\ \pm\lambda & \pm\gamma & \pm\omega\beta & \pm\mu\\ \nu & \beta & \gamma & \xi\\ \pm\omega\delta & 0 & 0 & \pm\alpha% \end{array}% \right) AB^{-1}, \]%
where% \[
B=\left( \begin{array}{ccc} \pm (\alpha\gamma -\omega\beta\delta ) & \pm (\omega\alpha\beta -\omega \gamma\delta ) & 0 \\ \alpha\beta -\gamma\delta & \alpha\gamma -\omega\beta\delta & 0 \\ \pm (\beta\lambda -\gamma\nu +\omega\beta\xi -\gamma\mu ) & \pm (\gamma \lambda -\omega\beta\mu +\omega\gamma\xi -\omega\beta\nu ) & \pm
(\gamma ^{2}-\omega\beta ^{2})% \end{array}% \right) . \]
We note that $\langle a,d\rangle +L^{2}$ is a characteristic subalgebra, and
first investigate the orbits of $pa,pd$. We consider three separate cases: $%
pa=pd=0$, $pa$ and $pd$ span a one dimensional subspace, and $pa,pd$ are
linearly independent. It turns out that there are $p+4$ orbits of $pa,pd$.
It is quite easy to see that if $pa,pd$ do \emph{not} span $\langle
ba,ca\rangle $ then we can assume that $pa=pd=0$, or $pa=0$, $pd=ca$, or $%
pa=0$, $pd=cb$, or $pa=ca$, $pd=cb$. There are $p$ orbits where $pa,pd$ span
$\langle ba,ca\rangle $, and we have a \textsc{Magma} program to find them.
\subsubsection{$pa=pd=0$}
If $pb,pc$ don't both lie in $\langle ba,ca\rangle $ then we can take $pb\in \langle ba,ca\rangle $ and $pc\notin\langle ba,ca\rangle $, which mean we
need to take $\beta =0$. We can then take $pc=cb$, which means we need to
take $\gamma =1$ in the $+$ matrices and $\gamma =-1$ in the $-$ matrices.
We can then take $pc=0$ or $ca$. There are $p$ orbits when $pb,pc\in\langle
ba,ca\rangle $, and there is a \textsc{Magma} program to find them.
\subsubsection{$pa=0$, $pd=ca$}
We need $\delta =0$, $\beta =0$ in both the plus and minus matrices, and $% \gamma =1$ in the plus matrices and $\gamma =-1$ in the minus matrices. We
then have:
So we can take $w=z=0$, and we can assume that $u=0,1$, or the least
non-square. (Experimentally only 0 and 1 arise, but I don't have a proof of
this.) There is a \textsc{Magma} program to find the orbits of $u,v,x,y$.
\subsubsection{$pa=ca$, $pd=cb$}
We need $\delta =0$, $\beta =0$ and $\gamma =1$ in both the plus and minus
matrices. You also need $\lambda =-\xi\omega $, $\mu =-\nu $, and $\alpha
=1 $. We then have:
If $pb,pc$ both lie in $\langle ba,ca\rangle $, then we can assume that $%
pb=pc=0$, and that $pa=ca$. There is a \textsc{Magma} program to find the $p$
orbits of $pd$.
If $pb,pc$ don't both lie in $\langle ba,ca\rangle $, then we can assume
that $pb=0$, and that $pc\in\langle ba,ca\rangle +cb$ though we then need $% \beta =0$, and $\gamma =1$ in the plus matrices and $\gamma =-1$ in the
minus matrices. This gives:
Note that the values of $pa$ and $pd$ depend only on $\alpha ,\delta $
(together with their original values), and that replacing $\alpha ,\delta $
by $\alpha k,\delta k$ makes no difference. There is a \textsc{Magma}
program to compute the orbits of $pa,pd$ under this action. It isn't
particularly easy to see, but for any fixed values of $pa,pd$, we can always
take $x=0$, and $y=0$ or 1. Just to make things tricky, for some fixed $%
pa,pd $, $x=y=0$ is in the same orbit as $x=0$, $y=1$, and sometimes it
isn't. There is a \textsc{Magma} program, notes4.1case6.m, to sort this out.
\end{document}
Messung V0.5
¤ Dauer der Verarbeitung: 0.41 Sekunden
(vorverarbeitet)
¤
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 und die Messung sind noch experimentell.