Classical edit

We review some terminology used in projective geometry.

A correlation of a projective geometry is a bijection on its subspaces that reverses containment. In particular, a correlation interchanges points and hyperplanes.^[2]

A correlation of order two is called a polarity.

A polarity is called a unitary polarity if its associated sesquilinear form s with companion automorphism α satisfies

s(u,v) = s(v,u)^α for all vectors u, v of the underlying vector space.

A point is called an absolute point of a polarity if it lies on the image of itself under the polarity.

The absolute points of a unitary polarity of the projective geometry PG(d,F), for some d ≥ 2, is a nondegenerate Hermitian variety, and if d = 2 this variety is called a nondegenerate Hermitian curve.^[3]

In PG(2,q²) for some prime power q, the set of points of a nondegenerate Hermitian curve form a unital,^[4] which is called a classical unital.

Let ${\displaystyle {\mathcal {H}}={\mathcal {H}}(2,q^{2})}$  be a nondegenerate Hermitian curve in ${\displaystyle PG(2,q^{2})}$  for some prime power ${\displaystyle q}$ . As all nondegenerate Hermitian curves in the same plane are projectively equivalent, ${\displaystyle {\mathcal {H}}}$  can be described in terms of homogeneous coordinates as follows:^[5]

Ree unitals edit

Another family of unitals based on Ree groups was constructed by H. Lüneburg.^[6] Let Γ = R(q) be the Ree group of type ²G₂ of order (q³ + 1)q³(q − 1) where q = 3^2m+1. Let P be the set of all q³ + 1 Sylow 3-subgroups of Γ. Γ acts doubly transitively on this set by conjugation (it will be convenient to think of these subgroups as points that Γ is acting on.) For any S and T in P, the pointwise stabilizer, Γ_S,T is cyclic of order q - 1, and thus contains a unique involution, μ. Each such involution fixes exactly q + 1 points of P. Construct a block design on the points of P whose blocks are the fixed point sets of these various involutions μ. Since Γ acts doubly transitively on P, this will be a 2-design with parameters 2-(q³ + 1, q + 1, 1) called a Ree unital.^[7]

Lüneburg also showed that the Ree unitals can not be embedded in projective planes of order q² (Desarguesian or not) such that the automorphism group Γ is induced by a collineation group of the plane.^[8] For q = 3, Grüning^[9] proved that a Ree unital can not be embedded in any projective plane of order 9.^[10]

Unitals with ${\displaystyle n=3}$  edit

In the four projective planes of order 9 (the Desarguesian plane PG(2,9), the Hall plane of order 9, the dual Hall plane of order 9 and the Hughes plane of order 9.^[b]), an exhaustive computer search by Penttila and Royle^[11] found 18 unitals (up to equivalence) with n = 3 in these four planes: two in PG(2,9) (both Buekenhout), four in the Hall plane (two Buekenhout, two not), and so another four in the dual Hall plane, and eight in the Hughes plane. However, one of the Buekenhout unitals in the Hall plane is self-dual,^[12] and thus gets counted again in the dual Hall plane. Thus, there are 17 distinct embeddable unitals with n = 3. On the other hand, a nonexhaustive computer search found over 900 mutually nonisomorphic designs which are unitals with n = 3.^[13]

Isomorphic versus equivalent unitals edit

Since unitals are block designs, two unitals are said to be isomorphic if there is a design isomorphism between them, that is, a bijection between the point sets which maps blocks to blocks. This concept does not take into account the property of embeddability, so to do so we say that two unitals, embedded in the same ambient plane, are equivalent if there is a collineation of the plane which maps one unital to the other.^[10]

By examining the classical unital in ${\displaystyle PG(2,q^{2})}$  in the Bruck/Bose model, Buekenhout^[14] provided two constructions, which together proved the existence of an embedded unital in any finite 2-dimensional translation plane. Metz^[15] subsequently showed that one of Buekenhout's constructions actually yields non-classical unitals in all finite Desarguesian planes of square order at least 9. These Buekenhout-Metz unitals have been extensively studied.^[16][17]

The core idea in Buekenhout's construction is that when one looks at ${\displaystyle PG(2,q^{2})}$  in the higher-dimensional Bruck/Bose model, which lies in ${\displaystyle PG(4,q)}$ , the equation of the Hermitian curve satisfied by a classical unital becomes a quadric surface in ${\displaystyle PG(4,q)}$ , either a point-cone over a 3-dimensional ovoid if the line represented by the spread of the Bruck/Bose model meets the unital in one point, or a non-singular quadric otherwise. Because these objects have known intersection patterns with respect to planes of ${\displaystyle PG(4,q)}$ , the resulting point set remains a unital in any translation plane whose generating spread contains all of the same lines as the original spread within the quadric surface. In the ovoidal cone case, this forced intersection consists of a single line, and any spread can be mapped onto a spread containing this line, showing that every translation plane of this form admits an embedded unital.

Hermitian varieties are in a sense a generalisation of quadrics, and occur naturally in the theory of polarities.

Definition edit

Let K be a field with an involutive automorphism ${\displaystyle \theta }$ . Let n be an integer ${\displaystyle \geq 1}$  and V be an (n+1)-dimensional vector space over K.

A Hermitian variety H in PG(V) is a set of points of which the representing vector lines consisting of isotropic points of a non-trivial Hermitian sesquilinear form on V.

Representation edit

Let ${\displaystyle e_{0},e_{1},\ldots ,e_{n}}$  be a basis of V. If a point p in the projective space has homogeneous coordinates ${\displaystyle (X_{0},\ldots ,X_{n})}$  with respect to this basis, it is on the Hermitian variety if and only if :

${\displaystyle \sum _{i,j=0}^{n}a_{ij}X_{i}X_{j}^{\theta }=0}$ 

where ${\displaystyle a_{ij}=a_{ji}^{\theta }}$  and not all ${\displaystyle a_{ij}=0}$ 

If one constructs the Hermitian matrix A with ${\displaystyle A_{ij}=a_{ij}}$ , the equation can be written in a compact way :

${\displaystyle X^{t}AX^{\theta }=0}$ 

where ${\displaystyle X={\begin{bmatrix}X_{0}\\X_{1}\\\vdots \\X_{n}\end{bmatrix}}.}$ 

Tangent spaces and singularity edit

Let p be a point on the Hermitian variety H. A line L through p is by definition tangent when it is contains only one point (p itself) of the variety or lies completely on the variety. One can prove that these lines form a subspace, either a hyperplane of the full space. In the latter case, the point is singular.

Citations edit