Automorphism groups of designs with $\lambda=1$
Extracted source text ยท arXiv 1805.12091
\documentclass{amsart}
\usepackage{amsmath, amsthm}
\usepackage{amssymb, latexsym}
\usepackage{amsfonts}
\usepackage{graphicx}
\usepackage[dvipsnames]{xcolor}
\usepackage{isomath}
\usepackage{tkz-graph}
%%%%%%%%%%%%%%%%%%%
%produces nicer calligraphic letters
\usepackage{calrsfs}
\newtheorem{Theorem}[equation]{Theorem}
\newtheorem{theorem}[equation]{Theorem}
\newtheorem{Proposition}[equation]{Proposition}
\newtheorem{proposition}[equation]{Proposition}
\newtheorem{lemma}[equation]{Lemma}
\newtheorem{Lemma}[equation]{Lemma}
\newtheorem{definition}[equation]{Definition}
\newtheorem{definitions}[equation]{Definitions}
\newtheorem{Corollary}[equation]{Corollary}
\newtheorem{corollary}[equation]{Corollary}
\newtheorem{conjecture}[equation]{Conjecture}
\newtheorem{fact}[equation]{\hspace{-3pt}\rm}
\newtheorem{remark}[equation]{Remark}
\newtheorem{rema}[equation]{Remark}
\newtheorem{Remark}[equation]{Remark}
\newtheorem{Remarks}[equation]{Remarks}
\newtheorem{hypothesis}[equation]{Hypothesis}
\newtheorem{coro}[equation]{Corollary}
\newtheorem{prop}[equation]{Proposition}
\newtheorem{exam}[equation]{Example}
\newtheorem{Example}[equation]{Example}
\newtheorem{Examples}[equation]{Examples}
\newtheorem{notation}[equation]{Notation}
\newtheorem{problem}[equation]{Problem}
\numberwithin{equation}{section}
%redefined \Gamma to be upright
\DeclareMathSymbol{\upGamma}{\mathalpha}{operators}{0}
\begin{document}
\title[Automorphism groups of designs with $\lambda=1$]
{Automorphism groups of designs with $\lambda=1$
}
\author{William M. Kantor
}
\address{U. of Oregon,
Eugene, OR 97403
\ and \
Northeastern U., Boston, MA 02115}
\email{kantor@uoregon.edu}
\begin{abstract}
\vspace{-4pt}
If $G$ is a finite group and $k =q>2$ or $k=q+1$ for a prime power $q$
then, for infinitely many integers $v$, there is a
$2$--$(v,k,1)$-design ${\bf D}$ for which ${\rm Aut} {\bf D}\cong G$.
\end{abstract}
\maketitle
\section{Introduction
}
\label{Introduction}
Starting with Frucht's theorem on graphs \cite{Frucht}, there have been many papers proving that any finite group is isomorphic to the full automorphism group of
some specific type of combinatorial object.
Babai surveyed this topic \cite{Ba1}, and in \cite[p.~8]{Ba1} stated that
in \cite{Ba2}
he had proved that
2-designs with $\lambda=1$ are such objects when
$k =q>2$ or $k=q+1$ for a prime power~$q$.
(The case of Steiner triple systems was handled in \cite{Me}.)
The purpose of this note is to provide a proof of Babai's result\footnote{This theorem was proved before I knew of Babai's result.}:
\begin{theorem}
\label{design construction3}
Let $G$ be a finite group and $q$ a prime power.
\begin{itemize}
\item [\rm(i)]There are infinitely many integers $v $ such that there is a
$2$--$(v,q+1,1)$-design ${\bf D}$ for which ${\rm Aut} {\bf D}\cong G$.
\item [\rm(ii)]
If $q>2$ then there are infinitely many integers $v $ such that there is a
\mbox{$2$--$(v,q,1)$-design} ${\bf D}$ for which ${\rm Aut} {\bf D}\cong G$.
\end{itemize}
\end{theorem}
Parts of our proof mimic \cite[Sec.~5]{DK} and \cite[Sec.~4]{Ka}, but the present situation is much simpler.
We modify a small number of subspaces of a projective
or affine space in such a way that the projective
or affine space can be recovered from the resulting design by elementary
geometric arguments. Further geometric
arguments determine the automorphism group.
%\smallskip\smallskip
Section \ref{Concluding remarks} contains further properties of the design {\bf D} in the theorem,
some of which are needed in future research \cite{DoK}.
\medskip
\noindent
\emph{Notation}:
We use standard permutation group notation, such as
$x^\pi$ for the image of a point $x$ under a permutation $\pi$ and
$g^h=h ^{-1}gh $ for conjugation.
The group of automorphisms of a projective space $Y={\rm PG}(V)$ defined by a vector space $V$ is denoted by ${\rm P\upGamma L}(V)={\rm P\upGamma L}(Y)$; this is induced by the group ${\rm \upGamma L}(V)$ of
invertible semilinear transformations on $V$. Also ${\rm A\upGamma L}(V)$ denotes the
group of automorphisms of the affine space ${\rm AG}(V)$ defined by $V$.
\section{A simple projective construction}
\label{Construction}
Let $G$ be a finite group. For suitable $n$ let
$\Gamma$ be a simple, undirected, connected graph on $\{1,\ldots ,n\}$ such that
${\rm Aut} \Gamma \cong G$ and $G$ acts semiregularly on the vertices (as in \cite{Frucht}).
All sufficiently large $n$ satisfy these conditions; we always assume that $n\ge6$.
Let $K={\bf F}_q \subset F={\bf F}_{q^4}$, and let $\theta$ generate $F^*$.
Let $V_F$ be an $n$-dimensional vector space over $F$,
with basis $ v_1,\ldots ,v_n $.
View $G$ as acting on $V_F$, permuting
$\{ v_1,\ldots ,v_n \}$ as it does $\{1,\ldots ,n\}$.
View $V_F$ as a vector space $V$ over $K$.
If $Y$ is a set of points of ${\bf P}={\rm PG}(V)$
then $\langle Y\rangle$ denotes the smallest subspace
of {\bf P} containing $Y$.
We will modify the
point-line design ${\rm PG}_1(V)$ of ${\bf P} $, using nonisomorphic designs
$ \Delta _1$ and $\Delta _2 $
whose parameters are those of ${\rm PG}_1(K^4)={\rm PG}_1(3,q)$ but are not isomorphic to
that design, chosen so that ${\rm Aut} \Delta _1$
fixes a point (Proposition~\ref{needed designs}).
Our design $\bf D$ has
the set $\mathfrak P$ of points of {\bf P}
as its set of points. Most blocks of $\bf D$
are lines of {\bf P}, with the following exceptions involving some of
the subspaces $Fv, $ $0\ne v\in V$, viewed as subsets of $\mathfrak P$.
For orbit representatives $i$ and $ij$
of $G$ on the vertices and edges of $\Gamma$,
\begin{enumerate}
\item [(I)] replace the set of lines of ${\rm PG}_1(F v _i)$ by
a copy of
the set of blocks of $\Delta _1$,
subject only to the condition
\begin{itemize}
\item [$(\#)$] there are distinct blocks, neither of which is a line of {\bf P},
whose span in {\bf P} is ${\rm PG}_1(F v _i)$,
\end{itemize}
\noindent
and then apply all $g\in G$ to these sets of blocks in order to obtain the blocks in
${\rm PG}_1((F v _i)^g),$ $g\in G; $ and
\item [(II)] replace the set of lines of ${\rm PG}_1(F (v _i +\theta v_j)) $ by
a copy of
the set of blocks of
$\Delta _2,$
subject only to $(\#)$,
and then apply all $g\in G$ to these sets of blocks in order to obtain the
blocks in ${\rm PG}_1(F (v _i +\theta v_j) ^g),$ $g\in G. $
\end{enumerate}
\indent
We need to check that these requirements can be met.
(i) \emph{Satisfying} $(\#)$:
Let $\bar\Delta _s$ be an isomorphic copy of $\Delta _s$, $s=1$ or 2, whose set of points is
that of ${\rm PG}_1(F v)={\rm PG}_1(F v _i)$ or ${\rm PG}_1(F(v _i +\theta v_j))$.
Let $B_1$ and $B_2$ be any distinct blocks of $\bar\Delta _s$.
Choose any permutation $\pi$ of the points of ${\rm PG}_1(F v)$ such that the sets
$B_1^\pi$ and $B_2^\pi$ are not lines of ${\rm PG}_1(F v)$ and together
span ${\rm PG}_1(F v)$.
Using $\bar\Delta _s^\pi$ in place of $ \bar\Delta _s$ satisfies $(\#)$. (If $q+1\ge4$ then $B_2$ is not needed.)
(ii)
\emph{These replacements are well-defined}:
For (II), if $F (v _i +\theta v_j) ^g \cap F (v _i +\theta v_j) ^{g'}\ne 0 $ for some $g,g'\in G,$
then $v _{i ^{g'}} +\theta v_{j ^{g'}}\in F (v _{i ^{g}} +\theta v_{j ^{g}}) $.
Then either
$v _{i ^{g'}} =v _{i ^{g}}$ and $v _{j ^{g'}} =v _{j ^{g}}$,
or $v _{i ^{g'}} =\alpha \theta v _{j ^{g}}$ and $\theta v _{j ^{g'}} =\alpha v _{i ^{g}}$
for some $\alpha\in F^*$;
but in the latter case we obtain $1= \alpha \theta $ and $ \theta=\alpha $,
whereas $\theta$ generates $F^*$.
Thus, $v _{i ^{g'}} =v _{i ^{g}}$, so the semiregularly of $G$ on $ \{ 1,\ldots , n\} $ implies that
$g'=g$, as required.
It is trivial to see that ${\bf D} $ {\em is a design having the same parameters as
${\rm PG}_1(V)$.}
Clearly $G$ acts on the collection of subsets of $\mathfrak P$ occurring in (I) or (II):
we can view $G$ as a subgroup of both ${\rm Aut} {\bf D}$ and ${\rm PGL}(V)$.
We emphasize that the sets in {\rm(I)} and {\rm(II)} occupy a tiny portion of the
underlying projective space:
most sets $Fv$ are unchanged. More precisely, in view of the definition of
{\bf D}:
\begin{equation}
\label{block is line}
\begin{array}{llll}
\mbox{\em Every block of {\,\bf D} not contained in a set {\rm(I)} or {\rm(II)}
is a line of {\,\bf P}.}
\\
\mbox{\em Every line of {\,\bf P} not contained in set {\rm(I)} or {\rm(II)}
is a block of {\,\bf D}.}
\end{array}
\end{equation}
Nevertheless, we will distinguish between the \emph{lines of {\bf P}}
and the \emph{blocks of {\bf D}}, even when the blocks happen to be lines.
A {\em subspace} of {\bf D} is a set of points
that contains the block joining any pair of its points.
(Examples: (I) and (II) involve subspaces of {\bf D}.)
A {\em hyperplane} of {\bf D}
is a subspace of {\bf D} that meets every block but does not contain every point.
We need further notation:
\begin{equation}
\label{block&line}
\mbox{
\em Distinct $y,z\in \mathfrak P$ determine a block $yz$ of {\,\bf D} and a line
$\langle y,z\rangle$ of {\,\bf P}. \hspace{8pt} } \vspace{-4pt}
\end{equation}
%
\begin{equation}
\label{xyz}
\begin{array}{llll}
\mbox{\em For distinct $y,z\in \mathfrak P$ and $x\in \mathfrak P- yz,$ }
\vspace{1pt}
\\
\mbox{\em$\langle x|y,z\rangle =\bigcup\big\{xp\mid p\in y'z',
y'\in xy\hspace{-.5pt} -\hspace{-.5pt}\{x\}, z'\in xz\hspace{-.5pt} -\hspace{-.5pt}\{x\}, \hspace{-.5pt} \{y,z\}\hspace{-.5pt}\ne\hspace{-.5pt} \{y',z'\}\big\}$.}
\end{array}
\end{equation}
Here (\ref{xyz}) depends only on {\bf D} not on {\bf P}, which will allow us to
recover {\bf P} from {\bf D}.
\begin{Lemma}
\label{planes}
If $y,z\in \mathfrak P$ are distinct$,$
then there are more than $\frac{1}{2}|\mathfrak P|$ points $x\in \mathfrak P-yz $ such that
\begin{itemize}
\item [\rm(1)] $\langle x,y,z\rangle$ is a plane
of {\,\bf P} every line of which$,$ except possibly $ \langle y,z\rangle,$ is a block of
{\,\bf D}$,$
\item [\rm(2)] $ \langle x|y,z\rangle=\langle x,y,z\rangle ,$
\item [\rm(3)] if $yz \subseteq \langle x|y,z\rangle $ then $ \langle y,z\rangle=yz ,$
and
\item [\rm(4)] if $yz \not \subseteq \langle x|y,z\rangle $ then
$\langle y,z\rangle$ is the union of the pairs
$\{y_1,z_1\}\subset \langle x|y,z\rangle$ such that $y_1z_1\not\subseteq \langle x|y ,z \rangle.$
\end{itemize}
\end{Lemma}
\proof
Let
\begin{equation}
\label{x}
x\notin yz\cup \mbox{$ \bigcup\hspace{1pt}$}\big \{ \langle y,z,Fv\rangle \mid
\mbox{$Fv$ in (I) or (II)$ \big \}.$ }
\end{equation}
There are more than
${(q^{4n}-1)/(q-1)} - n^2(q^6-1)/(q-1) -(q+1)> \frac{1}{2}|\mathfrak P| $ such points $x$.
Clearly $\langle x,y,z\rangle$ is a plane of {\bf P}.
\smallskip
(1) Let $L\ne \langle y,z\rangle$ be a line of $\langle x,y,z\rangle$,
so $\langle x,y,z\rangle = \langle y,z, L \rangle$.
If $L$ is not a block of {\bf D} then, by (\ref{block is line}), $L$ is contained in some set $Fv$ in (I) or (II), so $x\in \langle y,z,L\rangle\subseteq \langle y,z, Fv\rangle$ contradicts (\ref{x}).
\smallskip
(2) By (1), $\langle x,y\rangle $ and $ \langle x,z\rangle $ are blocks of {\bf D}.
Let $\{y',z'\}$ be as in (\ref{xyz}). Then $\{y',z'\}\subset \langle x,y,z\rangle$
and $\langle y' ,z'\rangle \ne \langle y ,z \rangle $.
By (1), $ y' z'=\langle y' ,z'\rangle \subseteq \langle x,y,z\rangle$ and
$xp = \langle x,p\rangle \subseteq \langle x,y,z\rangle$ for each point $p$ of $\langle y' ,z'\rangle $.
Then $\langle x|y,z\rangle\subseteq \langle x,y,z\rangle$.
Each point of $\langle x,y,z\rangle$ lies in such a line $\langle x,p\rangle $;
since that line is a block by (1), $\langle x,y,z\rangle\subseteq \langle x|y,z\rangle $.
\smallskip
(3) If $ yz\ne \langle y,z\rangle$ then, by (\ref{block is line}),
$ yz$ lies in some set $Fv$ in (I) or (II).
By hypothesis and (2), $yz \subseteq \langle x|y,z\rangle\cap Fv=\langle x,y,z\rangle\cap Fv=\langle y,z\rangle$. Thus, $ yz= \langle y,z\rangle.$
\smallskip
(4) We have $yz\ne \langle y,z\rangle$ since
$ \langle y,z\rangle\subseteq \langle x,y,z\rangle= \langle x|y,z\rangle$ by (2).
By (\ref{block is line}), since $ \langle y,z\rangle$ is not a block it is contained in
some set $Fv$ in (I) or (II).
For any $\{y_1,z_1\}$
in (4) we have $\{y_1,z_1\}\subseteq \langle x|y,z\rangle= \langle x,y,z\rangle$ by (2),
and $y_1z_1\not\subseteq \langle x,y,z\rangle$,
so $\langle y_1,z_1\rangle$ is not a block of {\bf D} and hence
$\langle y_1,z_1\rangle= \langle y,z\rangle$ by (1).
On the other hand, consider an arbitrary pair
$\{y_1,z_1\} \subset \langle y,z\rangle \subset Fv$. Then
$y_1 z_1 \subset Fv$ by the definition of {\bf D}. Since $\langle y,z\rangle $ is not a block,
$y_1 z_1\not \subseteq \langle y,z\rangle=\langle x|y,z\rangle\cap Fv$ by (2),
so $y_1 z_1\not \subseteq \langle x|y,z\rangle$.
Thus, $\langle y,z\rangle$ is the union of the pairs $\{y_1,z_1\}$ in (4). \qed
\smallskip \smallskip
\noindent
\emph{Proof of} Theorem~\ref{design construction3}(i).
We first recover the lines of {\bf P} from {\bf D}.
For distinct $y,z\in \mathfrak P$,
use each $x\notin yz$ in Lemma~\ref{planes}(3) or (4)
in order to obtain, more than $ \frac{1}{2}|\mathfrak P| $ times,
the same set of points that must be $\langle y,z\rangle $.
\emph{We have now reconstructed all lines of {\,\bf P} as subsets of $\mathfrak P$}.
Then we have also recovered ${\bf P},$ $V,$ ${\rm \upGamma L}(V)$
and ${\rm P\upGamma L}(V)$, so that
${\rm Aut} {\bf D}$ is induced by a subgroup of~${\rm Aut} {\bf P}={\rm P\upGamma L}(V)$.
Any block of {\bf D} that is not a line of {\bf P}
spans a 2-space or 3-space of {\bf P} occurring in
some 3-space ${\rm PG}_1(F v)$ in (I) or (II),
and spans at least a 4-space of {\bf P} together with any block in any
${\rm PG}_1(F v')\ne {\rm PG}_1(F v)$.
Any two blocks of {\bf D} that are not lines of {\bf P} and lie in the same set in (I) or (II)
span at most a 3-space of {\bf P}; by $(\#)$ each set in (I) or (II) is spanned by two such blocks.
\emph{This recovers all subsets {\rm(I)} and {\rm(II)} of $\,\mathfrak P$}
from {\bf D} and {\bf P}. Moreover,
the fact that $\Delta_1\not\cong\Delta_2$ specifies which of these subspaces of {\bf D}
have type (I) (or (II)).
We next determine the $F$-structure of $V$ using {\bf D}.
We claim that \emph{the subgroup of ${ {\rm PGL}}(V)$ fixing each set in
{\rm(I)} or {\rm(II)} arises from
scalar multiplications by members of $F^*$.}
Clearly such scalar multiplications behave this way. Let $h\in { {\rm PGL}}(V)$
behave as stated. Let $\hat h\in { {\rm GL}}(V)$ induce $h$.
Then
$\hat h\colon xv_i\mapsto (xA_i)v_i$ for all $x\in F$ and a $4\times4$
invertible matrix $A_i$ over $K$.
If $ij$ is an edge of $\Gamma$ and $x\in F$, then
$(x(v_i+\theta v_j))^{\hat h}= (xA_i ) v_i+ ((x\theta)A_j ) v_j$ is in $F(v_i+\theta v_j)$, so
$(xA_i ) \theta = (x\theta)A_j $. Since $ij=ji$, also $(xA_j ) \theta = (x\theta)A_i $, so
$(x\theta\theta)A_i=((x\theta)A_j)\theta= (xA_i) \theta\theta$, and $A_i$ commutes with multiplication
by $\theta^2$. By Schur's Lemma, $xA_i = xa_i $
for all $x\in F$ and some
$a_i \in F^*$.
Then $xa_i\theta=x\theta a_j $,
so $a_i=a_j$.
Since $\Gamma $ is connected,
all $a_i$ are equal, proving our claim.
In particular, the field $F$ and the $F$-space $V_F$
can be reconstructed from {\bf D}.
Since {\rm Aut}{\bf D} normalizes $F^*$ by the preceding paragraph,
$ {\rm Aut} {\bf D}\le{\rm P\upGamma L}(V_F).$
We know that $G$ is inside both ${\rm Aut} {\bf D}$ and ${\rm PGL}(V)$.
Since the sets in (II) correspond to (ordered) edges of $\Gamma$,
$ {\rm Aut} {\bf D}$ induces $ {\rm Aut} \Gamma\cong G$ on the collection of sets in (I).
Let $h\in {\rm Aut} {\bf D}\le {\rm P\upGamma L}(V_F)$. Multiply $h$ by an element of $G$ in order to have~$h$ fix
all $Fv_i$. Let $\hat h\in { {\rm \upGamma L}}(V_F)$ induce $h$,
with associated field automorphism $\sigma\in {\rm Aut} F $.
For each $i$ we have $v_i^{\hat h}=a_iv_i$ for some ${a_i\in F^*}$.
Let $ij$ be an edge of $\Gamma$ and write $b=a_j/a_i$. As above,
${F (v _i +\theta v_j)^{\hat h}}= {F (a_i v _i +\theta ^\sigma a_jv_j)}=
{F(v _i + \theta ^{\sigma} b v_j) }$~and $F (\theta v _i +v_j)^{\hat h}
= {F (\theta ^\sigma a_iv _i +a_jv_j)}
= F ( v _i +\theta ^{-\sigma }b v_j)$
both have type (II), so
$ \theta^\sigma b= \theta^{\pm1} $ and $ \theta^{-\sigma}b= \theta^{\mp1} .$
Then ${b^2=1}$, $ \theta^\sigma = \pm\theta^{\pm1} $, and hence $\sigma=1$ and $b=1$
since $\theta$ generates $F^*$.
The connectedness of $\Gamma $ implies that
all $a_i$ are equal: $\hat h$ is scalar multiplication by $a_1$.
Since $h$ fixes $Fv_1$ it induces an automorphism of the subspace of {\bf D}
determined by $Fv_1$.
By (I) and our condition
on $\Delta _1$, $h$ fixes a point $K cv_1 $ of $Fv_1$.
Then $K cv_1 = (Kcv_1) ^h= K ca_1v_1 $, so
$a_1\in K$. Thus,
$h=1$ on ${\mathfrak P}$ and $ {\rm Aut} {\bf D} \cong G$.\qed
\section{A simpler projective construction}
\label{simpler projective}
We need a fairly weak result (Proposition~\ref{needed designs}) concerning designs with the parameters of ${\rm PG}_ 1(3,q)$.
We know of two published constructions for designs having those parameters,
due to Skolem \cite[p.~268]{Wi}
and Lorimer \cite{Lo}. However, isomorphism questions seem difficult using
their descriptions. Instead, we will use a method that
imitates \cite{Sh,Ka} (but which was hinted at by Skolem's idea).
Consider a hyperplane $X$ of ${\bf P}={\rm PG}(d,q)$, $d\ge3$;
we identify {\bf P} with ${\rm PG}_1(d,q)$.
Let $\pi $ be any permutation of the points of $X$. Define a geometry ${\bf D}_\pi $ as follows:
\vspace{-2pt}
\begin{itemize}
\item [] the set $\mathfrak P$ of points is the set of points of {\bf P}, and
\item []
blocks are of two sorts:
\begin{itemize}
\item [] the lines of {\bf P} not in $X$, and
\item [] the sets $L^\pi $ for lines $L \subset X$.
\end{itemize}
\end{itemize}
\vspace{-2pt}
Once again it is trivial to see that ${\bf D}_\pi $
is a design having the same parameters as {\bf P}.
Note~that $\pi $ has nothing to do with the incidences
between points and the blocks not in~$X$.
We have a hyperplane $X$ of ${\bf D}_\pi $ such that the blocks of ${\bf D}_\pi $
not in $X$ are lines of a projective space {\bf P} for which $\mathfrak P$ is the set of points.
We claim that \emph{the lines of this projective space can be recovered
from ${\bf D}_\pi $ and $X$.}
Namely, we have all points and lines of {\bf P}
not in $X$. For distinct $y,z\in X$ and $x\notin X$,
the set $\langle x|y,z\rangle$ in (\ref{xyz}) consists of the points of the plane $\langle x,y,z \rangle$
of {\bf P}, and $\langle x|y,z\rangle \cap X$ is the line $\langle y,z\rangle$.
We have now obtained all lines of the original projective space~{\bf P}, as claimed.
It follows that ${\rm Aut}{\bf D}_\pi \le {\rm Aut}{\bf P}.$
The symbol $X$ is ambiguous: it will now mean either a set of points or
a hyperplane of the underlying \emph{projective space} (as in the next result). It will not refer to $X$ together with a different set of lines produced by a permutation $\pi$.
\begin{prop}
\label{double cosets}
The designs
${\bf D}_\pi $ and ${\bf D}_{\pi '}$ are isomorphic by an isomorphism sending $X$ to itself if and only if ${\pi }$ and ${\pi '}$ are in the same
${\rm P\upGamma L}(X), {\rm P\upGamma L}(X)$ double coset in ${\rm Sym}(X)$.
Moreover$,$ the pointwise stabilizer of $X$ in ${\rm Aut}{\bf D}_\pi $ is transitive on the points outside of $X,$ and the stabilizer $({\rm Aut}{\bf D}_\pi )_X$ of $X$
induces ${\rm P\upGamma L}(X)\cap {\rm P\upGamma L}(X)^\pi $ on~$X$.%
\end{prop}
\proof
Let $g{\colon\!} {\bf D}_\pi \to{\bf D}_{\pi '}$ be such an isomorphism. We just saw that
{\bf P} is naturally reconstructible from either design. It follows that
$g$ is a collineation of {\bf P}; its restriction $\bar g$ to $X$ is in ${\rm P\upGamma L}(X)$.
If $L\subset X$ is a line of {\bf P} then $g$ sends the block
$L^\pi \subset X$ of ${\bf D}_\pi $ to a block $L^\pi {}^g\subset X$ of ${\bf D}_{\pi '}$.
Then $L^\pi {}^g{}^{\pi '^{-1}}$ is a line of {\bf P}, so that
$\pi \bar g \pi '^{-1}$ is a permutation of the points of the hyperplane $X$ of {\bf P}
sending lines to lines, and hence is an element $ h\in {\rm P\upGamma L}(X)$.
Thus, $\pi $ and $\pi '$ are in the same ${\rm P\upGamma L}(X), {\rm P\upGamma L}(X)$ double coset.
Conversely, if $\pi $ and $\pi '$ are in the same ${\rm P\upGamma L}(X), {\rm P\upGamma L}(X)$ double coset let $\bar g, h\in {\rm P\upGamma L}(X)$ with $\pi \bar g\pi '^{-1}= h $.
Extend $\bar g$ to $g\in {\rm Aut} {\bf P} $ in any way. We claim that $g$ is an isomorphism
$ {\bf D}_\pi \to{\bf D}_{\pi '}$. It preserves incidences between
blocks not in $X$ and points of {\bf P}
since $g\in {\rm Aut} {\bf P} $ and those incidences have nothing to do with $\pi$ and $\pi'$.
Consider an incidence $x\in B\subset X$
for a block $B$ of ${\bf D}_\pi $. Then $B = L^\pi $ for a line $L\subset X$.
Since $g\in {\rm Aut} {\bf P} $, $x^g\in B^g=B^{\bar g}=L^{\pi \bar g}=(L^{ h})^{\pi '}$,
which is a block of ${\bf D}_{\pi '}$, as required.
For the final assertion, the pointwise stabilizer of $X$ in ${\rm Aut}{\bf P}$
is in ${\rm Aut}{\bf D}_\pi $ by the definition of $ {\bf D}_\pi $.
We have seen that the group induced on $ X$ by
${\rm Aut}{\bf D}_\pi $ corresponds to the pairs $(\bar g , h )\in
{\rm P\upGamma L}(X)\times {\rm P\upGamma L}(X)$ satisfying $\pi \bar g\pi ^{-1}= h $.
\qed
\smallskip\smallskip
Note that there are many extensions $g$ of $\bar g$ since the designs
${\bf D}_\pi $ have many automorphisms inducing the identity on $X$.
Double cosets arise naturally in this type of result; compare \cite[Theorem 4.4]{Ka}.
Let $v_i=(q^i -1)/(q-1)$.
\begin{coro}
\label{numbers}
There are at least $v_d!/ v_{d+1}|{\rm P\upGamma L}(d,q)|^2$
pairwise nonisomorphic designs having the same parameters as {\bf P}.
\end{coro}
\proof
Fix $\pi $ in the proposition.
There are at most $v_{d+1}$ hyperplanes $Y$ of ${\bf D}_\pi $
(as in \cite[Theorem 2.2]{JT}).
By the proposition there are then at most $|{\rm P\upGamma L}(X)|^2$ choices for
$\pi ' $ such that ${\bf D}_\pi \cong{\bf D}_{\pi '}$ by an isomorphism sending $Y$ to $X$.
Since there are $v_d!$ choices for $\pi $ we obtain the stated lower bound.\qed
\rema \em
We describe a useful trick. \em
A transposition $\sigma$ and a $3$-cycle $\tau$ are in different
${\rm P\upGamma L}(d,q), {\rm P\upGamma L}(d,q)$ double cosets in ${\rm Sym}(N),$
$ N= (q^d-1)/(q-1),$ if $d\ge 3$ and we exclude the case $d=3,q=2$. \rm
For, if $\sigma g = h \tau$ with $g,h\in {\rm P\upGamma L}(d,q)$ then
$ g^{-1} h =g^{-1}\cdot \sigma g \tau ^{-1} =
{\sigma} ^{ g}{\tau} ^{-1} \in {\rm P\upGamma L}(d,q)$ fixes at least $N-5$ points,
and hence is 1 by our restriction on $d$, whereas ${\sigma} ^{ g}\ne {\tau} $.
\begin{prop}
\label{needed designs}
For any $q$ there are two designs having the parameters of ${\bf P}={\rm PG}_1(3,q)$ and not isomorphic to one another or to
${\bf P} ,$ for one of which the automorphism group fixes a point.
\end{prop}
\proof
If $q=2 $ then there are even such designs with trivial automorphism group \cite{CCW}.
(Undoubtedly such designs exist for all $q$.)
Assume that $q>2$. The preceding corollary and remark provide us with two nonisomorphic designs. It remains to deal with the final assertion constructively.
Let $\pi $ be a transposition $(x_1,x_2)$ of $X$. We will show that ${\bf D}_\pi $ behaves as stated.
First note that each $g\in {\rm Aut}{\bf D}_\pi \le {\rm Aut}{\bf P}$ fixes $X$.
For, suppose that $Y=X^g\ne X$ for some $g$. The blocks in $Y$ not in $X$ are lines
of {\bf P}. Then the same is true of the
blocks in $Y^{g^{-1}}=X$ not in $X^{g^{-1}}$. This contradicts the fact that $\pi $ sends all lines $\ne\langle x_1,x_2\rangle$ of {\bf P} inside $X$ and on $x$ to sets that are not lines of {\bf P}.
By Proposition~\ref{double cosets},
${\rm Aut}{\bf D}_\pi =({\rm Aut}{\bf D}_\pi)_X $
induces ${\rm P\upGamma L}(X)\cap {\rm P\upGamma L}(X)^\pi $ on $X$.
Let $\pi \bar g\pi ^{-1}= h $ for $\bar g, h\in {\rm P\upGamma L}(X)$.
Then $\bar g^{-1} h =\pi ^{\bar g} \pi ^{-1}$
is a collineation of $X$ that moves at most $2\cdot 2$ points of $X$
and hence fixes at least $(q^2+q+1) -2\cdot 2 > q+\sqrt q +1 $ points.
By elementary (semi)linear algebra, the only such collineation is 1,
so that $\bar g= h$ commutes with $\pi $ and hence fixes the line $\langle x_1,x_2\rangle$.
Then $\bar g$ also fixes a point of $X$ and hence of ${\bf D}_\pi $.
\qed
\rema\rm By excluding the possibilities $q\le8$ and $q$ prime in the previous section
we could have used nondesarguesian projective planes (and ${[F\colon\!K]=3}$).
\section{A simple affine construction}
\label{affine construction}
\emph{We now consider }Theorem~\ref{design construction3}(ii).
The proof is similar to that of Theorem~\ref{design construction3}(i).
That result handles the cases $q=3,$ 4 or 5, but we ignore this and only assume that $q>2$.
Let $G$ and $\Gamma$ be as in Section~\ref{Construction}.
This time we use $K={\bf F}_q\subset F={\bf F}_{q^3}$; once again $\theta$ generates $F^*$.
Let $V_F$ be an $n$-dimensional vector space over $F$,
with basis $ v_1,\ldots ,v_n $.
View $V_F$ as a vector space $V$ over $K$.
If $Y$ is a set of points of {\bf A} then $\langle Y\rangle$ denotes the smallest affine subspace
containing $Y$.
We will modify the
point-line design $ {\rm AG}_1(V)$ of ${\bf A}= {\rm AG}(V)$, using nonisomorphic designs
$ \Delta _1,\Delta _2 $
whose parameters are those of $ {\rm AG}_1(3,q)$ but are not isomorphic to
that design, chosen so that ${\rm Aut} \Delta _1$ fixes at least two points
(Proposition~\ref{needed resolvable designs}).
Our design $\bf D$ has $V$ as its set of points. Most blocks of $\bf D$
are lines of {\bf A}, with exceptions involving the
sets $Fv, $ $0\ne v\in V$, in Section~\ref{Construction}(I,\,II),
where now {\em $Fv$ is viewed as a $3$-dimensional affine space.}
As before, the set of lines of
$ {\rm AG}_1(Fv_i)$ or $ {\rm AG}_1(F (v _i +\theta v_j))$
is replaced by a copy of the set of blocks of $\Delta_1$ or $\Delta_2$.
This time, for each of these we require
\begin{itemize}
\item [$(\#')$] there are distinct blocks, each of which spans a plane of {\bf A},
such that the intersection of those planes is a line.
\end{itemize}
Clearly, these two blocks span a 3-space.
(When $q>3$ it would be marginally easier to require that there is a single block
that spans a 3-space.)
Condition $(\#')$ can be satisfied
exactly as in \emph{Satisfying} $(\#)$ in Section~\ref{Construction}.
Since different sets $Fv $ meet only in a single point, the modifications made
inside them are unrelated.
Once again it is easy to check that this produces a design {\bf D}
with the desired parameters for which $G\le {\rm Aut} {\bf D}$.
As in Section~\ref{Construction},
most sets $Fv$ are unchanged.
In view of the definition of {\bf D}, the analogue of (\ref{block is line}) holds.
We use the natural analogues of definitions (\ref{block&line}) and (\ref{xyz}),
using {\bf A} in place of {\bf P} and $V$ in place of $\mathfrak P$.
\begin{Lemma}
\label{aplanes}
If $y,z\in V$ are distinct$,$
then there are more than $\frac{1}{2}|V|$ points $x\in V-yz $ such that
\begin{itemize}
\item [\rm(1)] every line of the plane
$\langle x,y,z\rangle$ of {\bf A}$,$ except possibly $ \langle y,z\rangle,$ is a block of {\bf D}$,$
\item [\rm(2)] $ \langle x|y,z\rangle =\langle x,y,z\rangle ,$
\item [\rm(3)] if $yz \subseteq \langle x|y,z\rangle $ then $ \langle y,z\rangle=yz ,$
and
\item [\rm(4)] if $yz \not \subseteq \langle x|y,z\rangle $ then
$\langle y,z\rangle$ is the union of the pairs
$\{y_1,z_1\}\subset \langle x|y,z\rangle $ such that $y_1z_1\not\subseteq \langle x|y ,z \rangle .$
\end{itemize}
\end{Lemma}
\proof
Using $x$ in (\ref{x}), this is proved exactly as in Lemma~\ref{planes} except for (2), where
we need to consider parallel lines using blocks that are lines by (1).
Clearly $ \langle x|y,z\rangle \subseteq \langle x,y,z \rangle$;
we must show that $ \langle x,y,z\rangle \subseteq \langle x|y,z\rangle.$
In (\ref{xyz}), for $p$ in the line
$y' z ' =\langle y' ,z '\rangle$ of $\langle x,y,z\rangle$ parallel to $\langle y,z\rangle$,
the blocks $xp\subset \langle x|y,z\rangle$ cover all points
of the plane $\langle x,y,z\rangle$ except for those in the line $L$ on $x$ parallel
to $\langle y,z\rangle$. If $y'\in xy-\{x,y\}$ and
$p'=y'z\cap L$, then $L=xp'\subset \langle x|y,z\rangle $,
so $ \langle x,y,z\rangle \subseteq \langle x|y,z\rangle.$ \qed
\smallskip \smallskip
\noindent\emph{Proof of\,\,\,}Theorem~\ref{design construction3}(ii).
First recover all lines of {\bf A} from {\bf D} exactly
as in the proof of Theorem~\ref{design construction3}(i).
This also produces both the $K$-space $V$ and ${\rm A\upGamma L}(V)$ from {\bf D}.
We recover all subsets (I) and (II) essentially as before.
Consider a pair $B,B'$ of blocks of {\bf D} behaving as in $(\#')$: $\langle B\rangle $
and $\langle B'\rangle $ are planes and $\langle B\rangle \cap \langle B'\rangle $ is a line.
Since distinct subsets in (I) or (II) do not have a common line, each such pair
$B,B'$ spans a subset in (I) or (II).
Thus, by $(\#')$ we have obtained each subset in (I) or (II) from {\bf D} and {\bf A} using
some pair $B,B'$. Once again, the fact that $\Delta_1\not\cong\Delta_2$
specifies which of these subspaces of {\bf D} have type (I) (or (II)).
The subsets (I) all contain $0$, and ${\rm Aut} {\bf D}$ fixes their intersection,
so ${\rm Aut} {\bf D}$ is induced by a subgroup of ${\rm A\upGamma L}(V)_0
={ {\rm \upGamma L}}(V )$.
Recover the field $F$ exactly as in the proof of Theorem~\ref{design construction3}(i).
Once again, $ {\rm Aut} {\bf D}$ is a subgroup of ${\rm \upGamma L}(V_F)$ that
induces $ {\rm Aut} \Gamma\cong G$ on the collection of sets in (I).
By repeating the argument at the end of the proof of Theorem~\ref{design construction3}(i) we
reduce to the case of $ h\in {\rm Aut} {\bf D}$
fixing all sets in (I) and acting on $V$ as $v\mapsto av$ for some
$a\in F^*$. We chose $\Delta_1$ so that $ {\rm Aut} \Delta_1$ fixes at least two of its points. It follows that $a=1$, so that $h=1$ and $ {\rm Aut} {\bf D}\cong G$. \qed
\section{A simpler affine construction}
\label{simpler affine}
Consider a plane $X$ of
${\bf A}= {\rm AG}(3,q)= {\rm AG}(V)$, $q>2$; we identify {\bf A} with $ {\rm AG}_1(3,q)$.
Let $\pi $ be any permutation of the points of $X$.
Define a geometry ${\bf D}_\pi $ as follows:
\vspace{-2pt}
\begin{itemize}
\item [] the set $V$ of points is the set of points of {\bf A}, and
\item []
blocks are of two sorts:
\begin{itemize}
\item [] the lines of {\bf A} not in $X$, and
\item [] the sets $L^\pi $ for lines $L\subset X$.
\end{itemize}
\end{itemize}
\vspace{-2pt}
Once again it is trivial to see that ${\bf D}_\pi $ is a design having the same parameters as {\bf A}.
As in Section~\ref{simpler projective}, the blocks of ${\bf D}_\pi $
not in $X$ are lines of an affine space {\bf A} for which $V$ is the set of points.
As in Sections~\ref{simpler projective} and \ref{affine construction}, \emph{the lines of this affine space can be recovered
from ${\bf D}_\pi $} using the analogue of (\ref{xyz}).
\begin{prop}
\label{double cosets 2}
The designs
${\bf D}_\pi $ and ${\bf D}_{\pi '}$ are isomorphic by an isomorphism sending $X$ to itself if and only if ${\pi }$ and ${\pi '}$ are in the same
${\rm A\upGamma L}(X), {\rm A\upGamma L}(X)$ double coset in ${\rm Sym}(X)$.
This produces at least $q^2!/q(q^2+q+1)| {\rm A\upGamma L}(2,q) | ^2$ pairwise
nonisomorphic designs having the same parameters as ${\rm AG}_1(3,q) $.
Moreover$,$ the pointwise stabilizer of $X$ in ${\rm Aut}{\bf D}_\pi $ is transitive on the points outside of $X,$
and $({\rm Aut}{\bf D}_\pi)_X $
induces ${\rm A\upGamma L}(X)\cap {\rm A\upGamma L}(X)^\pi $ on $X$.
\end{prop}
\proof
This is the same as for Proposition~\ref{double cosets} and Corollary~\ref{numbers}.
\qed
\begin{prop}
\label{needed resolvable designs}
For any $q \ge3$ there are at least two designs having the parameters of ${\bf A}= {\rm AG}_1(3,q),$
not isomorphic to one another or to
${\bf A} ,$ such that the automorphism group of one of them fixes at least two points.
\end{prop}
\proof
The bound in the preceding proposition provides us with many nonisomorphic designs.
We need to deal with the requirement concerning automorphism groups.
By \cite{LR} we may assume that $q\ge4$.
Let $\pi\in {\rm Sym}(X)$ be a 4-cycle $ (x,x_1,x_2,x_3)$, where
$x_1,x_2,x_3$ are on a line not containing $x$.
We will show that ${\bf D}_{\pi}$ behaves as required.
Let $g\in {\rm Aut}{\bf D}_{\pi }$. As in the proof of Proposition~\ref{needed designs},
$g$ fixes $X$ and induces a collineation $\bar g$ of the subspace $X$ of {\bf A}.
By Proposition~\ref{double cosets 2}, $\pi \bar g= h \pi $ with $\bar g, h\in {\rm A\upGamma L}(X)$.
As before, $\bar g^{-1} h =\pi ^{\bar g}\pi ^{-1}$ is a collineation of $X$ that
fixes at least $q^2 -2\cdot 4> q $ points as $q \ge 4$.~Then
$\bar g= h$ and $ \pi ^{\bar g}=\pi $.
Since the collineation $\bar g$ commutes~with~$\pi $
it fixes $ \{x,x_1,x_2,x_3\}$ and hence also $x$, and so
is the identity on the support of $\pi $. Thus, $ {\rm Aut}{\bf D}_{\pi }$ is the identity on that support. \qed%
\section{Steiner quadruple systems}
\label{Steiner quadruple systems}
We have avoided ${\rm AG}(d,2)$ in the preceding two sections.
Here we briefly comment about those spaces in the context of $3$--$(v,4,1)$-designs
(Steiner quadruple systems), outlining a proof of the following result in \cite{Me}.
\begin{theorem}
\label{SQS}
If $G$ is a finite group then there are infinitely many integers $v $ such that there is a
$3$--$(v,4,1)$-design ${\bf D}$ for which ${\rm Aut} {\bf D}\cong G$.
\end{theorem}
\proof
Let $K={\bf F}_2\subset F= {\bf F}_{16}$ be as in Section~\ref{Construction}, with $\theta $
a generator of $F^*$. Let $V_F $ be a vector space with basis
$v_1,\dots,v_n$, viewed as a $K$-space $V$. This time we modify the 3-design
${\rm AG}_2(V)$ of points and (affine) planes of $V$. We use
nonisomorphic designs $\Delta_1, \Delta_2$ having the parameters of ${\rm AG}_2(4,2)$
but not isomorphic to that design, and such that ${\rm Aut}(\Delta_1)=1$ \cite{KOP}.
Once again our design $\bf D$ has $V$ as its set of points. Most blocks of $\bf D$
are planes of {\bf A}, with exceptions involving the
sets $Fv, $ $0\ne v\in V$, in Section~\ref{Construction}(I,\,II),
where now {\em $Fv$ is viewed as a $4$-dimensional affine space.}
As before, the set of planes of
$ {\rm AG}_2(Fv_i)$ or $ {\rm AG}_2(F (v _i +\theta v_j))$
is replaced by a copy of the set of blocks of $\Delta_1$ or $\Delta_2$.
This time, for each of these we require
\begin{itemize}
\item [$(\#'')$] there are distinct blocks, each of which spans a 3-space of {\bf A},
such that the intersection of those 3-spaces is a plane.
\end{itemize}
Once again it is easy to check that this produces a design {\bf D}
with the desired parameters for which $G\le {\rm Aut} {\bf D}$.
{Distinct $x,y,z\in V$ determine a block $xyz$ of {\,\bf D} and a plane
$\langle x,y,z\rangle$ of {\,\bf A}. }
For distinct $x,y,z$ and $w\notin xyz,$ instead of (\ref{xyz}) we use
{$\langle w|x,y,z\rangle\!=\bigcup\big\{abc\mid
a\in wxy-\{w\}, b\in wxz-\{w\}, c\in wyz-\{w\}$, with $a,b,c$ distinct and not all in $\{x,y,z\}\big\}$.}
As before, all planes of {\bf A} can be recovered from {\bf D}, this time using
various sets $\langle w|x,y,z\rangle$. Also the sets in (I) and (II) can be recovered,
as can $F$, and
the argument at the end of Section~\ref{affine construction} goes through as before. \qed
\section{Concluding remarks}
\label{Concluding remarks}
\Remark\rm
\label{additional}
When considering possible consequences of this paper it became clear that additional properties of our designs should also be mentioned.
\smallskip \smallskip
\begin{enumerate}
\item [\rm(1)]
Additional properties of the design {\bf D}
in Theorem~\ref{design construction3}(i).
\begin{enumerate}
\item[\rm(a)]
{\em ${\rm PG}(3,q)$-connectedness}.
The following graph is connected: the vertices are the subspaces of {\bf D}
isomorphic to ${\rm PG}_1(3,q)$, with two joined when they meet.
\item[\rm(b)]
{\em ${\rm PG}(n-1,q)$ generation}.
{\bf D} is generated by its subspaces isomorphic to ${\rm PG}_1(n-1 ,q)$.
\item[\rm(c)]
Every point of {\bf D} is in a subspace isomorphic to
${\rm PG}_1(n-1 ,q)$ (in fact, many of these).
\item[\rm(d)] More than $q^n$ points are moved by every nontrivial automorphism of {\bf D}.
\end{enumerate}
\item [\rm(2)]
Additional properties of the design {\bf D} in Theorem~\ref{design construction3}(ii).
\begin{enumerate}
\item[\rm(a)]
{\em ${\rm AG}(3,q)$-connectedness}.
The following graph is connected: the vertices are the subspaces of {\bf D}
isomorphic to ${\rm AG}_1(3,q)$, with two joined when they meet.
\item[\rm(b)]
{\em ${\rm AG}(n ,q)$ generation}.
{\bf D} is generated by its subspaces isomorphic to ${\rm AG}_1(n ,q)$.
\item[\rm(c)]
Every point of {\bf D} is in a subspace isomorphic to
${\rm AG}_1(n ,q)$ (in fact, many of these).
\item[\rm(d)] More than $q^n$ points are moved by every nontrivial automorphism of {\bf D}.
\end{enumerate}
\item [\rm(3)]
Additional properties of the design {\bf D}
in Theorem~\ref{SQS}. This time versions of (2a) (using ${\rm AG}_2(4,2)$-connectedness),
(2b), (2c) and (2d) hold.
\end{enumerate}
\smallskip \smallskip
These reflect the fact that the sets of points in (I) or (II) cover a tiny portion of the
underlying projective or affine space: a subset of the points determined by $F$-linear
combinations of at most two of the $v_i$.
For (1a), it is easy to see that any point in $\mathfrak P$ lies in a 4-space of $V$
that contains some point $ K \beta \sum_i\!v_i$, $\beta\in F^*$,
and meets each set in (I) or (II) in at most a point; by (\ref{block is line}) this produces
a subspace of {\bf D} isomorphic to ${\rm PG}_1(3,q)$. Moreover,
all $K \beta \sum_iv_i$ lie in $ F( \sum_i\!v_i)$, which also produces
a subspace of {\bf D} isomorphic to ${\rm PG}_1(3,q)$.
For (1b) we give examples of subspaces of $V$:
\noindent
$\langle v_1 +\theta ^2v_2, v_2+\theta ^2v_3+\theta ^iv_4, \dots ,v_{n-2}+\theta ^2v_{n-1}+\theta^i v_n, v_1+v_2+v_4+v_5, \theta(v_1+v_2+v_4+v_5)\rangle$
\noindent
for $2< i<q^4-1$. Each of these misses all sets in (I) or (II), and hence
determines a subspace of {\bf D} isomorphic to
${\rm PG}_1(n-1 ,q)$. These subspaces generate a subspace of {\bf D} containing
the points $K(\theta^i-\theta^3)v_n$, $3< i<q^4-1$, and hence also
${\rm PG}_1(Fv_n)$. Now permute the subscripts to generate {\bf D}.
Part (1c) holds by using $K$-subspaces similar to the above ones.
There are clearly projective spaces of larger dimension that are subdesigns of {\bf D}.
Part (1d) depends on the semiregularity of $G$ on $\{ v_1,\ldots ,v_n \}$. Use the points
$K \sum_i\alpha _i v_i$ with $\alpha_1=1$ and $\alpha_i\in F-\{1\}$ for $i>1$,
where each $\alpha \in F-\{1\}$ occurs either for 0 or at least two basis vectors $v_i$.
The lower bound $q^n$ is easy to obtain but very poor.
Both (2) and (3) are handled as in (1).
\Remark \rm
In (II) we used the $K$-subspaces $F (v _i +\theta v_j)$. We could have used
subspaces $F (v _i +\theta_rv_j)$, $r=1,\dots,s,$ for various $\theta_r$, together with
further nonisomorphic designs $\Delta_{2,r}$
(which are needed to distinguish among the $F (v _i +\theta_r v_j)$). All proofs go through without difficulty,
as do the additional properties in the preceding remark.
\Remark \rm
Each of our designs has the same parameters as
some $ {\rm PG}_1(V)$ or $ {\rm AG}_1(V)$.
What is needed is a much better type of result, such as:
\emph{for each finite group $G$ there is an integer $f(|G|)$ such that$,$
if $q$ is a prime power and if $v>f(|G|)$ satisfies the necessary conditions for
the existence of a $2$--$(v,q+1,1)$-design$,$ then there is such a design {\bf D}
for which ${\rm Aut}{\bf D}\cong G$.}
When $q=2$ this result is proved in a sequel to the present paper \cite{DoK}.
\smallskip \smallskip \smallskip \smallskip
{\noindent\em Acknowledgements.} \rm
I am grateful to Jean Doyen for providing me with a clear description of Skolem's construction and for helpful comments concerning this research.
I am also grateful to a referee for many helpful comments.
This research was supported in part by a grant from the Simons Foundation.
\begin{thebibliography}{GMC}
\bibitem[Ba1]{Ba1}
L. Babai,
On the abstract group of automorphisms, in:
Combinatorics (Swansea, 1981). LMS Lecture Notes 52, Cambridge U. Press, Cambridge-New York 1981, pp. 1--40.
\bibitem[Ba2]{Ba2}
L. Babai, BIBD's with given automorphism groups, unpublished; see \cite[p.~8]{Ba2}.
\bibitem[CCW]{CCW}
F.~N. Cole, L.~D. Cummings, H.~S. White, The complete enumeration of triad systems in 15 elements. Proc. Natl. Acad. Sci. 3 (1917) 197--199.
\bibitem[DK]{DK}
U. Dempwolff, W.~M.~Kantor, Distorting symmetric designs,
Des. Codes Cryptogr. 48 (2008) 307--322.
\bibitem[DoK]{DoK}
J.~Doyen, W.~M.~Kantor, Automorphism groups of Steiner triple systems.
http://arxiv.org/abs/1808.03615
\bibitem[Fr]{Frucht}
R. Frucht, Herstellung von Graphen mit vorgegebener
abstrakter Gruppe, Compositio Math. 6 (1938) 239--250.
\bibitem[JT]{JT}
D.~Jungnickel, V. ~D. ~Tonchev, The number of designs with geometric parameters grows exponentially, Des. Codes Cryptogr. 55 (2010) 131--140.
\bibitem[Ka]{Ka} W.~M.~Kantor,
Automorphisms and isomorphisms of symmetric and affine designs,
J. Alg. Comb. 3 (1994) 307--338.
\bibitem[KOP]{KOP}
P. Kaski, P. \"Osterg\aa rd, O. Pottonen,
The Steiner quadruple systems of order 16,
J. Combin. Theory Ser. A 113 (2006) 1764--1770.
\bibitem[LR]{LR}
C.~C. Lindner, A. Rosa,
On the existence of automorphism
free Steiner triple systems,
J. Algebra 34 (1975) 430--443.
\bibitem[Lo]{Lo}
P. Lorimer, A class of block designs having the same parameters as the design of points and lines in a projective 3-space, in:
Combinatorial mathematics
(Proc. Second Australian Conf., Univ. Melbourne, Melbourne, 1973). Lecture Notes in Math. 403, Springer, Berlin~1974, pp. 73--78 .
\bibitem[Me]{Me}
E. Mendelsohn, On the groups of automorphisms of
Steiner triple and quadruple systems, J. Combin. Theory Ser. A 25 (1978) 97--104.
\bibitem[Sh]{Sh} S.~S. Shrikhande, On the nonexistence of affine resolvable balanced incomplete block designs, Sankhy$\bar {\rm a}$ 11 (1951) 185--186.
\bibitem[Wi]{Wi}
E. Witt, \"Uber Steinersche Systeme, Abh. Math. Sem. Hamburg 12 (1938) 265--275.
\end{thebibliography}
\end{document}