Global Methods for Combinatorial Isoperimetric Problems

The failure of compression, or any combinatorial alternative, to solve the EIP on
or the VIP on
, ultimately forced combinatorialists back to the old workhorse of classical mathematics: calculus. By this we mean passing to a continuous limit, solving the continuous problem by variational means and arguing that the essence of the combinatorial problem is preserved in the limit.
If we let n 1 = n 2 = = n d = n ? ? , then
, the d-dimensional unit cube. Given any ideal, S ? [0, 1] d, with the product partial order, then
the ( d-dimensional) Lebesgue measure of S. Also, if ? ( S) is the area of the free boundary of S (that part of the boundary of S in the interior of [0, 1] d), measured wrt the L 1 (taxicab) metric, then
Thus the analog of the EIP on [0, 1] d is, given v ? [0, 1], to minimize ? ( S) over all ideals S ? [0, 1] d such that ? ( S) = v. We claim that the only critical (locally optimal) stable ideals for the EIP on [0, 1] d are S ( j, d, t) = [0, 1] d ? j [0, t] j, 1 ? j