Normal spreads (Q1301604)

From MaRDI portal
This is the item page for this Wikibase entity, intended for internal use and editing purposes.
Please use this page instead for the normal view: Normal spreads
scientific article; zbMATH DE number 1334358
Language Label Description Also known as
default for all languages
No label defined
    English
    Normal spreads
    scientific article; zbMATH DE number 1334358

      Statements

      Normal spreads (English)
      0 references
      0 references
      11 September 2000
      0 references
      A \((t - 1)\)-spread in \(\Sigma = PG(n - 1, q)\) is a family \({\mathcal S}\) of mutually disjoint subspaces of dimension \(t - 1\) which partition the points of \(\Sigma\). A necessary and sufficient condition for the existence of a spread is \(t |n\), and so we assume \(n = rt\) for some integer \(r\) from now on. There is a natural \(2 - (q^{rt}, q^t, 1)\) translation design with parallelism associated with each \((t - 1)\)-spread, which is a translation plane when \(r = 2\). If \(r > 2\), a \((t - 1)\)-spread \({\mathcal S}\) is called normal (or geometric) if \({\mathcal S}\) induces a spread in any subspace generated by two of its elements. The associated design \(A({\mathcal S})\) is isomorphic to \(AG(r,q^t)\) if and only if \({\mathcal S}\) is normal. Normal spreads have many applications in finite geometry from the construction of packings to the classification of certain kinds of flag-transitive affine planes. R. C. Bose pointed out that taking the elements of a normal \((t - 1)\)-spread \({\mathcal S}\) as ``points'', the subspaces of dimension \(2t - 1\) joining two elements of \({\mathcal S}\) as ``lines'', and the natural inclusion as ``incidence'' yields an isomorphic copy of \(PG(r - 1, q^t)\). In the paper under review the author represents normal spreads by using Grassmannians. Let \(\Sigma = PG(rt - 1, q)\), and let \({\mathcal G} \subseteq \overline \Sigma = PG ( {{rt}\choose{t}} - 1, q)\) be the Grassmannian of \((t - 1)\)-subspaces of \(\Sigma\). Let \(g\) be the associated Grassmannian map. If \({\mathcal S}\) is a normal \((t - 1)\)-spread of \(\Sigma\), it is shown here that there exists an \((r^t - 1)\)-subspace \(\Delta\) of \(\overline \Sigma\) such that \(g ({\mathcal S}) = \Delta \cap {\mathcal G}\) is an algebraic variety \({\mathcal V}_{r,t}\) which is a \((q^{t(r - 1)} + q^{t(r - 2)} + \ldots + q^t + 1)\)-cap of \(\Delta \cong PG(r^t - 1, q)\). Moreover, the collineation group of \(\Delta\) stabilizing \({\mathcal V}_{r,t}\) acts 2-transitively on its points. As a corollary it is shown that \({\mathcal V}_{r,2}\) is the union of 3-dimensional elliptic quadrics, and for \(q\) even \({\mathcal V}_{r,3}\) is the union of ovoids of the hyperbolic quadric \(Q^+ (7,q)\). Now restrict to the case \(r = 3\) and \(t = 2\), so that the Bose model for \(PG(2,q^2)\) is obtained from a normal 1-spread \({\mathcal S}\) of \(PG(5,q)\) as previously described. Let \({\mathcal T}\) be the lines of \({\mathcal S}\) corresponding to some Bose subplane of \(PG(2, q^2)\). Then it is shown that there is a 5-dimensional subspace \(S\) of \(\Delta \cong PG(8,q)\) such that \(g({\mathcal T}) = S \cap {\mathcal V}_{3,2}\) is a Veronesean surface of \(S\). From this it follows that \({\mathcal V}_{3,2}\) is partitioned into \(q^2 - q + 1\) Veronesean surfaces. Similarly, it is shown that a Hermitian curve of \(PG(2,q^2)\) is represented on the variety \({\mathcal V}_{3,2}\) by the intersection of \({\mathcal V}_{3,2}\) with a certain hyperplane of \(\Delta\). If \(q \equiv 0,2 \pmod 3\), this hyperplane section of \({\mathcal V}_{3,2}\) is a ``unitary'' ovoid of \(Q^+ (7,q)\) [see \textit{W. M. Kantor}, Can. J. Math. 34, 1195-1207 (1982; Zbl 0493.51006) for an alternate description]. The final ingredient of this paper has to do with blocking sets. Let \(r = 3\) and consider the Bose model for \(\Pi = PG(2, q^t)\) obtained from a normal \((t - 1)\)-spread \({\mathcal S}\) of \(\Sigma = PG (3t - 1, q)\). If \(L\) is a \(t\)-subspace of \(\Sigma\) not contained in one of the \((2t - 1)\)-subspaces joining two elements of \({\mathcal S}\), it is shown that the lines of \({\mathcal S}\) meeting \(L\) correspond to a nontrivial blocking set of \(\Pi\). Such blocking sets are called linear. Many known blocking sets, including those of Rédei-type, turn out to be linear. A class of non-Rédei minimal blocking sets, which are also linear, has recently been constructed in [\textit{P. Polito} and \textit{O. Polverino}, Combinatorica 18, No. 1, 133-137 (1998; Zbl 0910.05017)]. This idea is then extended to construct a blocking set of \(PG(3,q^2)\) (i.e., meets every line but contains no hyperplane) of the smallest possible size.
      0 references
      normal spreads
      0 references
      Grassmannians
      0 references
      caps
      0 references
      ovoids
      0 references
      blocking sets
      0 references
      0 references

      Identifiers