Some Bounds for the Number of Blocks III

Abstract

Let $\mathcal D=(\Omega, \mathcal B)$ be a pair of $v$ point set $\Omega$ and a set $\mathcal B$ consists of $k$ point subsets of $\Omega$ which are called blocks. Let $d$ be the maximal cardinality of the intersections between the distinct two blocks in $\mathcal B$. The triple $(v,k,d)$ is called the parameter of $\mathcal B$. Let $b$ be the number of the blocks in $\mathcal B$. It is shown that inequality ${v\choose d+2i-1}\geq b\{{k\choose d+2i-1} +{k\choose d+2i-2}{v-k\choose 1}+....$ $.+{k\choose d+i}{v-k\choose i-1} \}$ holds for each $i$ satisfying $1\leq i\leq k-d$, in the paper: Some Bounds for the Number of Blocks, Europ. J. Combinatorics 22 (2001), 91--94, by R. Noda. If $b$ achieves the upper bound, $\mathcal D$ is called a $\beta(i)$ design. In the paper, an upper bound and a lower bound, $ \frac{(d+2i)(k-d)}{i}\leq v \leq \frac{(d+2(i-1))(k-d)}{i-1} $, for $v$ of a $\beta(i)$ design $\mathcal D$ are given. In the present paper we consider the cases when $v$ does not achieve the upper bound or lower bound given above, and get new more strict bounds for $v$ respectively. We apply this bound to the problem of the perfect $e$-codes in the Johnson scheme, and improve the bound given by Roos in 1983.