1: %auto-ignore
2: \paragraph{Abstract.}\footnote{A part of this research was
3: completed when this author held a position at the
4: Institute for System Analysis, Moscow.}
5: Let $\Si^\ep_n$ be the graph defined on the (+)-points of an %FDC
6: $n$-dimen\-sio\-nal
7: GF(3)-space carrying a nondegenerate symmetric bilinear form with
8: discriminant $\ep$, points are adjacent iff they are perpendicular. We prove
9: that if $\ep=1$, $n\geq 6$ (resp. $\ep=-1$, $n\geq 7$)
10: then $\Si^\ep_{n+1}$ is the unique connected locally
11: $\Si^\ep_n$ graph. One may view this result as a characterization of a
12: class of $c^k\cdot C_2$-geometries (or 3-transposition groups).
13: We briefly discuss an application of the result to a %FDC
14: characterization of
15: Fischer's sporadic groups.
16: %
17: \section{Introduction and results}
18: The study of geometries of classical groups as point-line
19: systems with fixed local structure is very extensive. For instance, see
20: Tits~\cite{Ti:la}. We refer the reader to a paper of Cohen and Shult~\cite{CoSh}
21: for a
22: brief survey on more recent results.
23:
24: In this paper we use the aforementioned approach to characterize some
25: ``nonclassical" geometries, namely 3-transposition graphs that arise %FDC
26: from
27: orthogonal GF(3)-groups.
28:
29: For the concept of a %FDC
30: 3-transposition graph
31: and a 3-transposition group, %FDC
32: see Fischer~\cite{Fi}. In
33: case of characteristic 2 groups such graphs come from classical geometries.
34: It is worthwhile to mention %FDC
35: the work of Hall and Shult~\cite{HaSh}
36: characterizing some class of 3-transposition graphs as {\it locally
37: cotriangular} ones. Note that the graphs $\Si^\ep_n$ considered in the
38: present paper are not locally cotriangular (with finitely %FDC
39: many
40: exceptions for small $n$). %FDC misprint exeptions
41:
42: There is a one-to-one correspondence between 3-transposition
43: graphs and {\it Fischer spaces} (see, e.g. Buekenhout \cite{Bue:fi}, Cuypers
44: \cite{Cuy:genfisp}, Weiss \cite{We1}, \cite{We2}), namely 3-transposition graphs
45: are complements to the collinearity graphs of Fischer spaces. It turns
46: out that it is possible to exploit this duality and classification of
47: Fischer spaces in the final part of the proof of our theorem~\ref{o3n:mainth}.
48: Note, however, that locally $\Si^\ep_n$ graphs not always arise from
49: Fischer spaces. E.g., there are at least two nonisomorphic
50: locally $\Si^-_6$-graphs, one of them has diameter 4, (i.e. it does not
51: correspond to any Fischer space) see e.g. \cite{BCN}. The author has %FDC
52: shown
53: [in preparation] that they are the only examples of such graphs.
54:
55: Throughout the paper we consider undirected graphs without loops and multiple
56: edges.
57: Given a graph $\Ga$, let us denote the set of vertices by
58: $\V=\V\Ga$, the set of edges by $\E=\E\Ga$.
59: Given two graphs $\Ga,\/\Delta$, the graph $\Ga\cup\Delta$
60: (resp. the graph $\Ga\cap\Delta$) is
61: the graph with the vertex set $\V\Ga\cup \V\Delta$ (resp.
62: $\V\Ga\cap \V\Delta$) and the edge set $\E\Ga\cup \E\Delta$
63: (resp. $\E\Ga\cap \E\Delta$). Given $v\in \V\Ga$, we denote by
64: $\Ga_i(v)$ the subgraph induced by vertices at distance $i$ from $v$,
65: and $\Ga_1(v)=\Ga(v)$.
66: Furthermore, $\Ga(X)=\bigcap_{x\in X} \Ga(x)$.
67: To simplify the notation we use
68: $\Ga(v_1,\dots,v_k)$ instead of $\Ga(\{v_1,\dots,v_k\})$ and
69: $u\in\Ga_i(\dots)$ instead of $u\in \V\Ga_i(\dots)$.
70: As usual, $v=v(\Ga)=|\V\Ga|$, $k=k(\Ga)=v(\Ga(x))$, where
71: $x\in\V\Ga$. Let $y\in\Ga_2(x)$. We denote
72: $\mu=\mu(x,y)=\mu(\Ga)=v(\Ga(x,y))$. Of course, %FDC misprint cource
73: we use
74: $k$, $\mu$ if it makes sense, %FDC
75: i.e. if those numbers are independent on the
76: particular choice of the corresponding vertices.
77: If $\Delta$ is a (proper) subgraph of
78: $\Ga$ we denote this fact as $\Delta\subseteq\Ga$ (resp.
79: $\Delta\subset\Ga$).
80:
81: We denote the complete multipartite graph with the $m$ parts of equal size $n$
82: by $K_{n\times m}$.
83: $\Aut(\Ga)$ denotes the automorphism group of $\Ga$.
84: Our group-theoretic notation is as in \cite{Atl}.
85: Let $\Ga$, $\Delta$ be two graphs. We say that $\Ga$ is
86: {\it locally} $\Delta$ if $\Ga(v)\cong\Delta$ for any $v\in \V(\Ga)$.
87: Let $\Ga,\ \overline\Ga$ be two graphs. We say that $\Ga$ is a
88: {\it cover} of $\overline\Ga$ if there exists a mapping $\varphi$
89: from $\V\Ga$ to $\V\overline\Ga$ which maps edges to edges.
90: Suppose we have a chain of graphs $\Si_1,\dots,\Si_n$, such that
91: $\Si_i$ is locally $\Si_{i-1},\ i=2,\dots,n$. Then for any complete
92: $k$-vertex subgraph $\Up$ of $\Si_m$, $1<k<m\leq n$
93: we have $\Si_m(\V\Up)\cong\Si_{m-k}$.
94: We say that a graph $\Ga$ is a {\it triple} graph if for each nonedge
95: $(u,v)$ there exist a unique $w\in \V\Ga$ such that
96: $\Ga(u,v)=\Ga(u,w)=\Ga(v,w)$.
97:
98: We slightly adopt notation and several basic facts from \cite{Mei}.
99: Let $T=T_n$ be an %FDC
100: $n$-dimensional GF(3)-vector space carrying a nondegenerate
101: symmetric bilinear form (,) with discriminant $\ep$. We say that the point
102: $\<v\>\subset T$ (or a nonzero vector $v\in T$) is of type $(+)$, $(-)$ or
103: {\it isotropic} according to $(v,v)=1,\, -1,\, 0 $ respectively. Since the form
104: is
105: constant on a point, the notation like $(p,q)$ for points $p$, $q$ will be
106: used freely. The orthogonal complement of $X\subseteq T$ in $T$ is denoted by
107: $X^\perp$.
108:
109: Define the graph $\Si^\ep_n=\Ga(\V,\E)$ as follows. Let $\V$ be the
110: set of (+)-points of $T$. Define $\E=\{(u,v)\subset \V\times\V |
111: (u,v)=0\}$, i.e. the edges are pairs of perpendicular (+)-points.
112: Given $\Ga=\Si^\ep_n$, we denote by $T(\Ga)$ the underlying
113: GF(3)-space. Note that $\Si^\ep_n$, $n\geq 5$ if $\ep=1$, $n\geq 4$ if
114: $\ep=-1$, is a rank 3 graph with automorphism group
115: $GO^\mu_n(3)$, where Witt defect $\mu$ is empty if $n$ is odd, otherwise
116: $\mu=(-\ep)^{n/2}$.
117: %FDC the use of the notation $\mu$ here is maybe badly chosen,
118: %FDC moreover, although
119: %FDC I know what you mean, I am not sure that everyone likes a sentence as
120: %FDC "$\mu$ is empty if $n$ is odd".
121: Since $T_n$ can be represented as the orthogonal direct sum of a (+)-point
122: and $T_{n-1}$, the graph $\Si^\ep_n$ is locally $\Si^\ep_{n-1}$.
123: We denote $\Si^1_n$ by $\Si^+_n$, and $\Si^{-1}_n$ by $\Si^-_n$.
124: Note that $\Si^\ep_n$ is a triple graph.
125: We refer the interested reader to \cite{Mei} for more detailed
126: information about $\Si^\ep_n$.
127: We will prove the following theorem. %FDC
128:
129: \begin{thm} Let $\Th=\Th_{n+1}$ be a connected locally $\Si^\ep_{n+1}$-graph,
130: where $\ep=1$, $n\geq 6$, or $\ep=-1$, $n\geq 7$.
131: Then $\Th$ is isomorphic to $\Si^\ep_{n+1}$.
132: \label{o3n:mainth}
133: \end{thm}
134:
135: \subsubsection*{\bf Remark 1} %FDC
136: One may view the graphs $\Th_{n+1}$ as the collinearity graphs
137: of %FDC
138: certain $c^k\cdot C_2(s,t)$-geometries ${\cal G}(\Th_{n+1})$
139: (here $k=n-(\ep+9)/2$), i.e. rank $k+2$ geometries
140: with diagram
141: $$\circ \!{\phantom{00000}\over \phantom{0}}
142: \!\!\circ\cdots\circ \!\!{\phantom{00000}\over \phantom{0}}
143: \!\!\circ \!\!{{\phantom{00} c \phantom{00}}\over \phantom{0}}
144: {\vphantom{\bigl\langle}}_s
145: \!\!\!\circ \!\! = \!\! = \!\! = \!\! = \!\! =
146: \!{\vphantom{\bigl\langle}}_t\!\!\!\circ$$
147: Conversely, the elements of the geometry may be viewed as $i$-cliques
148: of the graph with natural incidence, $i=1,2,\dots, k+1, s+k+1$.
149: Meixner has proved the following result~\cite{Mei}. %FDC
150:
151: \begin{res}
152: Let ${\cal G}$ be a residually connected flag-transitive
153: $c^k\cdot C_2$-geometry. Then if the $c^1\cdot C_2$-residues (resp.
154: $c^3\cdot C_2$-residues) of ${\cal G}$ are isomorphic to ${\cal G}(\Si^+_6)$
155: (resp. to ${\cal G}(\Si^-_7)$) then ${\cal G}={\cal G}(\Si^+_{k+5})$
156: (resp. ${\cal G}={\cal G}(\Si^-_{k+4}))$.
157: \label{o3n:Meires}
158: \end{res}
159:
160: Our theorem %FDC
161: implies that %FDC
162: the flag-transitivity
163: assumption can be replaced to a geometric condition (X) from \cite{Mei}. (X)
164: states that in the collinearity graph of ${\cal G}$ each $i$-clique is the
165: shadow of some element, $i=1,2,\dots k+1,s+k+1$.
166:
167: %\medskip %FDC
168: \subsubsection*{\bf Remark 2}%FDC
169: The significance of $\Si^+_{i+3}$ as subgraphs of
170: 3-transposition graphs $\Delta_{2i}$ of Fischer's sporadic simple groups %FDC
171: $Fi_{2i}$ ($i=2,3,4$) is well-known.
172: Namely, for $\Ga=\Delta_{2i}$ the subgraph $\Ga(x,y)$, where $x$ and $y$ are two
173: vertices at distance 2, is isomorphic to $\Si^+_{i+3}$. The author
174: used theorem~\ref{o3n:mainth} %FDC
175: to prove the following result~\cite{Pa:fi}. %FDC
176:
177: \begin{res}
178: Any connected locally $\Delta_{22}$ (resp.\ $\Delta_{23}$) graph is isomorphic
179: to $\Delta_{23}$ (resp. to $\Delta_{24}$ or to its 3-fold antipodal cover).
180: \qed
181: \end{res}
182:
183: \section{Proof of the theorem} %FDC
184:
185: \paragraph{Preliminaries} A proof of the following technical statement is omitted.
186:
187: \begin{lem}
188: Let $\Ga=\Si^\ep_n$, $n\geq 5$,
189: and let $a,b,c$ be isotropic %FDC
190: points of $T=T(\Ga)$. \hfil\break
191: \indent {If $(a,b)=0$, $a\not=b$, $(a,c)\not=0$ then} \hfil\break
192: \indent {(i) $a^\perp\cap c^\perp\cap\Ga\cong\Si^{-\ep}_{n-2}$,} \hfil\break
193: \indent {(ii) $a^\perp\cap b^\perp\cap\Ga$ is not isomorphic to
194: $\Si^{-\ep}_{n-2}$.} \hfil\break
195: \indent {(iii) Denote $\Up_p=\Ga\cap p^\perp$. If
196: $\Up_a\cap\Up_b$ contains $\{v\}\cup\Up_a(v)$ for some
197: $v\in\V\Up_a$ then $a=b$.}
198: \hfil\break
199: \indent {(iv) For any $u\in\V\Ga\setminus\V\Up_a$, the subgraph
200: $\Ga(u)\cap\Up_a$ is isomorphic to $\Si^{-\ep}_{n-2}$. }
201: %\qed %FDC
202: \label{L4}
203: \end{lem}
204:
205: \paragraph{Neighbourhood of two vertices at distance two}
206: We start with a simple general fact. Let $\Delta$ be a connected
207: graph satisfying the following property
208:
209: \begin{description}
210: \item[(*)] For any $u\in \V\Delta$ and
211: $v\in \V\Delta\setminus (\V\Delta(u)\cup\{ u \})$
212: the subgraph $\Delta(u,v)$ is isomorphic to some $M_{\Delta}$, whose
213: isomorphism type is independent on the particular choice of $u$ and $v$.
214: \end{description}
215:
216: \begin{lem}
217: Let $\Ga$ be locally $\Delta$ graph, where $\Delta$
218: satisfies {\rm (*)}. Then for %FDC
219: any $u\in \V\Ga$, $v\in\Ga_2(u)$
220: the graph $\Ga(u,v)$ is locally $M_{\Delta}$. \qed
221: \label{L6}
222: \end{lem}
223:
224: Note that $\Ga=\Si^\ep_n$ satisfies (*), when $\ep=1$, $n\geq 5$
225: (resp. $\ep=-1$, $n\geq 4$). Indeed, the stabilizer
226: of $u\in \V\Ga$ in $\Aut(\Ga)$ acts transitively on
227: $\Ga_2(u)$. This implies (*).
228: Thus lemma~\ref{L6} %FDC
229: holds for locally $\Ga$ graphs. The next statement
230: characterizes locally $M_\Ga$-subgraphs of $\Ga$.
231:
232: \begin{pr}
233: Let $\Ga=\Si^\ep_n$, and either $\ep=1$, $n\geq 6$,
234: or $\ep=-1$, $n\geq 7$. Let $\Om\subset\Ga$ be a locally $M_\Ga$ graph.
235: Then there exists a unique isotropic point $p\subset T(\Ga)$ such that
236: $\Om=\Ga\cap p^\perp$.
237: \label{P7}
238: \end{pr}
239:
240:
241: \begin{pf}
242: We proceed by induction on $n$. It is straightforward to check its basis.
243: We leave it to the reader, noticing that for
244: $\ep=1$, $n=6$ (resp. $\ep=-1$, $n=6$) it suffices to classify locally
245: $K_{3\times 2}\cup K_{3\times 2}\cup K_{3\times 2}$
246: (resp. $K_{3\times 4}\cup K_{3\times 4}\cup K_{3\times 4}$) subgraphs of
247: $\Ga$.
248:
249: Now let us check the inductive step. Let $\Om$ be a connected component
250: of a locally $M_\Ga$ subgraph of $\Ga$, $v_1\in\V\Om$,
251: $v_2\in\Om(v_1)$. By the inductive hypothesis,
252: $\Om(v_i)=\Ga(v_i)\cap p^\perp_i$, where $p_i$ is an
253: isotropic point of $T(\Ga(v_i))$, hence of $T(\Ga)$ ($i=1,2$).
254: Since $p_2\subset v^\perp_1$, it defines a locally $M_{\Ga(v_1)}$
255: subgraph $\Up$ of $\Ga(v_1)$, and $\Up(v_2)=\Om(v_1,v_2)$.
256: Hence by lemma~\ref{L4}~(iii), %FDC
257: applied to $\Ga(v_1)$, $p_1=p_2$.
258: Therefore $\Om=\Ga\cap p^\perp_1$. Finally, it is easy to check that
259: $\Om$ is a unique connected component of the subgraph under consideration.
260: \end{pf}
261:
262: \paragraph{Final part of the proof} Let $\Th$ be a connected locally
263: $\Ga=\Si^\ep_n$ graph. Here we assume either $\ep=1$, $n\geq 6$ or
264: $\ep=-1$, $n\geq 7$. Pick a vertex $u\in\V\Th$.
265:
266: \begin{lem}
267: \hfil\break
268: \indent (i) $\mu(\Th)=\mu(\Si^\ep_{n+1})$, $v(\Th)=v(\Si^\ep_{n+1})$.\hfil\break
269: \indent (ii) $\Th$ is a triple graph.
270: \label{L8}
271: \end{lem}
272:
273: \begin{pf}
274: (i). The first claim follows from proposition~\ref{P7}.
275: Indeed, counting in two ways
276: the edges between $\Th(u)$ and $\Th_2(u)$, we obtain the precise value of the
277: number of vertices of $\Th$. Let $v\in\Th_2(u)$, $w\in\Th(v)\setminus\Th(u)$.
278: By lemma~\ref{L4}~(iv) we obtain $\Th(u,v,w)\cong\Si^{-\ep}_{n-2}$, hence
279: nonempty.
280: Thus the diameter of $\Th$ equals two, and we are done.
281:
282: (ii). Assume that there exist three distinct vertices $v_i\in\Th_2(u)$,
283: $i=1,2,3$ such that $\Up=\Th(u,v_1)=\Th(u,v_2)=\Th(u,v_3)$.
284: It contradicts the fact that $\Th(w)$, where
285: $w\in\V\Up$, is a triple graph.
286: Observe that $|\Th_2(u)|$ is exactly twice the number of isotropic points
287: in $T(\Ga)$. Hence we have no choice determining the edges between $\Th(u)$
288: and $\Th_2(u)$. Now since $\Si^\ep_{n+1}$ is a triple graph, the same is
289: true for %FDC
290: $\Th$.
291: \end{pf}
292:
293: Let us denote $\Ga=\Th(u)$, $\Xi=\Th_2(u)$. Let $\Up$ be the graph defined
294: on isotropic points of $T(\Ga)$, two points are adjacent if they are
295: not perpendicular.
296:
297: \begin{pr}
298: $\Xi$ is a two-fold cover of $\Up$.
299: \label{P9}
300: \end{pr}
301:
302: \begin{pf}
303: Let $(v_1,v_2)\in\E\Xi$. By lemma~\ref{L4}~(iv) we have $\Th(u,v_1,v_2)\cong
304: \Si^{-\ep}_{n-2}$. Denote by $p_i$ the isotropic point of $T(\Ga)$
305: such that $p_i^\perp\cap\Ga=\Th(u,v_i)$ ($i=1,2$). By proposition~\ref{P7} such
306: a point
307: $p_i$ exists and is unique ($i=1,2$). By lemma~\ref{L4}~(i),
308: (ii), $(p_1,p_2)\not=0$.
309:
310: Conversely, assume $p_1$, $p_2$ are nonperpendicular isotropic points of
311: $T(\Ga)$. Denote by $\Om_i$ the locally $M_\Ga$ subgraph of
312: $\Ga$, which corresponds to $p_i$ ($i=1,2$).
313: For each $\Om_i$ we have exactly two vertices
314: $v_{ij}\in\Th_2(u)$ such that $\Om_i=\Th(u,v_{ij})$ ($i,j=1,2$).
315: Considering the neighbourhood of $v_{11}$, we see that
316: both $v_{21}$ and $v_{22}$ cannot be ajacent to $v_{11}$.
317: On the other hand $k(\Xi)=k(\Up)$. Hence one of $v_{21}$ and $v_{22}$ must
318: be ajacent to $v_{11}$. We have shown that the mapping
319: $v\mapsto\Th(u,v)$, where $v\in\V\Xi$, is a covering from $\Up$ to $\Xi$.
320: \end{pf}
321:
322: The latter statement implies that $\Xi$ possesses %FDC misprint posesses
323: an involutory automorphism
324: $g_u$ which interchanges any $v,w\in\V\Xi$ such that $\Th(u,v)=\Th(u,w)$, and
325: fixes $\{u\}\cup\Th(u)$ pointwise.
326:
327: Consider the subgroup $G_u$ of $\Aut(\Th)$ generated by $g_x$, $x\in\Th(u)$.
328: We have $G_u\cong\Aut(\Ga)$. Therefore $\Th$ is the collinearity graph of a
329: $c^k\cdot C_2$-geometry satisfying the conditions of
330: result~\ref{o3n:Meires}. %FDC
331: Hence our result follows from it. However, we would like to give
332: a complete proof of theorem~\ref{o3n:mainth} here.
333:
334: Our claim is that $\{g_v|v\in\V\Th\}$ is a class of 3-transpositions in
335: $\Aut(\Th)$. Indeed, clearly for any $x\in\Th(u)$ the involutions $g_u$ and
336: $g_x$ commute. Now let $y\in\Th_2(u)$. We must prove $\tau=(g_u g_y)^3=1$.
337: Note that $\tau$ belongs to the kernel of the
338: action of the stabilizer of every $v\in\Th(u,y)$ on $\Th(u)$. Therefore
339: it fixes every vertex of $\Th$.
340: Our claim is proved. The use of the classification of 3-transposition groups
341: given in \cite{Fi} completes the proof of theorem~\ref{o3n:mainth}.
342: \qed %FDC
343:
344:
345: \subsubsection*{\bf Note} %FDC
346: Hans Cuypers (personal communication) has suggested another idea how to
347: complete the proof, which is much more geometric. Namely, it may be
348: easily
349: shown that the partial linear space on $\V\Th$, whose lines are triples,
350: is an irreducible Fischer space (see introduction) %FDC
351: (or a locally polar
352: geometry with affine planes, see Cuypers and Pasini \cite{CuPa}). Then
353: the use of classification of these objects \cite{Cuy:genfisp} (resp.
354: \cite{CuPa}) completes the proof.
355:
356: