Spreading grid cells

Abstract

Let $S$ be a set of $n^2$ symbols. Let $A$ be an $n\times n$ square grid with each cell labeled by a distinct symbol in $S$. Let $B$ be another $n\times n$ square grid, also with each cell labeled by a distinct symbol in $S$. Then each symbol in $S$ labels two cells, one in $A$ and one in $B$. Define the \emph{combined distance} between two symbols in $S$ as the distance between the two cells in $A$ plus the distance between the two cells in $B$ that are labeled by the two symbols. Belén Palop asked the following question at the open problems session of CCCG 2009: How to arrange the symbols in the two grids such that the minimum combined distance between any two symbols is maximized? In this paper, we give a partial answer to Belén Palop's question. Define $c_p(n) = \max_{A,B}\min_{s,t \in S} \{\dist_p(A,s,t) + \dist_p(B,s,t) \}$, where $A$ and $B$ range over all pairs of $n\times n$ square grids labeled by the same set $S$ of $n^2$ distinct symbols, and where $\dist_p(A,s,t)$ and $\dist_p(B,s,t)$ are the $L_p$ distances between the cells in $A$ and in $B$, respectively, that are labeled by the two symbols $s$ and $t$. We present asymptotically optimal bounds $c_p(n) = Θ(\sqrt{n})$ for all $p=1,2,...,\infty$. The bounds also hold for generalizations to $d$-dimensional grids for any constant $d \ge 2$. Our proof yields a simple linear-time constant-factor approximation algorithm for maximizing the minimum combined distance between any two symbols in two grids.