# Generalised quadrangles V: elation generalised quadrangles

The focus of this post is elation generalised quadrangles. These are generalised quadrangles defined with respect to certain automorphisms of the generalised quadrangle. My main sources for this post have been Michel (Celle) Lavrauw’s and Maska Law’s Phd theses which are both availiable from Ghent’s PhD theses in finite geometry page.

Let be a generalised quadrangle of order with a point . An *elation* about is an automorphism of that fixes , fixes each line incident with and fixes no point not collinear with . If there exists a group of order consisting entirely of elations and which acts regularly on the set of points not collinear with then is called an *elation generalised quadrangle* or simply an EGQ. The group is called an *elation group* and is called the *base point*.

Many of the known examples of generalised quadrangle are EGQs. For example, consider defined with respect to the alternating form where

Let be the point given by the 1-dimensional subspace spanned by . Now contains the subgroup

Then and acts trivially on , so fixes each line incident with . Moreover, the points not collinear with are all points of the form such that . The group acts regularly on this set and so each element of is an elation and is an EGQ. When is even, is abelian while when is odd is known as a Heisenberg group. In particular, for a prime is an extraspecial group of exponent .

Of all the examples we have seen so far the only generalised quadrangles which are not EGQ’s and nor are their duals, are the Payne-derived quadrangles of order .

We should note here a couple of points:

- Not every point of an EGQ need be a base point. Indeed it was shown by Koen Thas and Hendrik van Maldeghem in 2008 that if every point of an EGQ is a base point then the EGQ is either a classical GQ or the dual of a classical GQ.
- It was recently noticed by Stan Payne and Koen Thas that the set of all elations about of an EGQ does not necessarily form a group.
- It was proved by Robert Rostermundt that for even has two nonisomorphic elation groups.
- It is not known if the elation group of a EGQ need be a -group but it is conjectured to be so.

** Kantor families **

In 1980 Bill Kantor discovered a general way of constructing EGQs as coset geometries as follows:

Let be a group of order with . Let be a family of subgroups of of order and be a family of subgroups of order such that for each , . We then define an incidence structure with points

- (i) the elements of ,
- (ii) the right cosets ,
- (iii) a point ;

and lines

- (a) the right cosets ,
- (b) the symbols .

Incidence is defined as follows: a point of type (i) is incident with each line , a point of type (ii) is incident with the line and each line contained in , and the point is incident with each line .

The incidence structure just defined is a GQ of order if and only if

- for distinct ,
- for .

A collection of such subgroups , in a group yielding a GQ is called a *(Kantor) 4-gonal family*.

The group acts as a group of automorphisms of our GQ by fixing the symbols and , and acting on the remaining points and lines by right multiplication. In particular, each element of is an elation about and so we have constructed an EGQ.

Not only does a Kantor 4-gonal family give rise to an EGQ, but each EGQ can be constructed in such a manner. We now outline this construction.

Let be an EGQ with elation group and base point . Since acts regularly on the set of points not collinear with , this set of points can be identified with the elements of . Let be a line incident with . By the GQ property, there is a unique line incident with the point and meeting . Let be the point of intersection of and . Now let be another line not incident with but meeting and let be the point of intersection. Let be a point not collinear with and incident with . Then and since is an EGQ, fixes . By the GQ property, is the unique line incident with that meets and is the unique line incident with that meets . Thus and . In particular, acts transitively on the set of lines not incident with that meet . There are such lines, so and the set of such lines can be identified with the cosets of in . We also have that acts transitively on the set of points other than that are incident with . There are such points and so and the set of such points can be indentified with the cosets of in . Now and is identified with the line , which is incident with . Moreover, a point identified with , for will be incident with the line and note that is contained in . Running over all lines incident with we obtain subgroup of order and subgroups of order , and all lines not incident with have been identified with a coset of a subgroup of order and each point collinear with has been identified with a coset of a subgroup of order . Finally, we identify with the symbol and each line incident with with the symbol .

It remains to check that the collection of subgroups we have obtained satisfy the two conditions that guarantee that Kantor’s construction yields a GQ. Let for . Since , it follows that the coset is identified with a line incident with . Moreover, the point identified with is incident with this line. Since , we also have that the point identified with is incident with . Hence is a line through meeting at . By the GQ property there is a unique line through meeting and so .

Finally, consider . If then is the line through the point of that meets . In particular, is the set of all points not collinear with that lie on a line that meets both and . Hence if then is a point of that also lies on a line meeting both and . Now from the way we constructing the Kantor family, the point lies on and . Thus if then the point also lies on the line that meets . Hence we have two lines through ( and ) that meet , contradicting the GQ property. Hence .

I mentioned in my last post on flock generalised quadrangles that given a -clan it is possible to construct a generalised quadrangle of order via a group coset construction. This is done by constructing a group and using the -clan to construct the two families and of subgroups of . I will not go through this here but it is outlined in both Maska’s and Celle’s theses.

The next post will be the last in the series and will be on translation generalised quadrangles. Unlike this post it will include some GQs not seen so far.