We show that if SAT does not have small circuits, then there must exist a small number of formulas such that every small circuit fails to compute satisfiability correctly on at least one of these formulas. We use this result to show that if P_{}^{NP} [1] = P_{}^{NP[2]}, then the polynomial-time hierarchy collapses to S_2^P \subseteq \sum {} _2^P \cap \prod {} _2^P Even showing that the hierarchy collapsed to \sum {} _2^P remained open prior to this paper.
Citation:
Lance Fortnow, A. Pavan, Samik Sengupta, "Proving SAT does not have Small Circuits with an Application to the Two," ccc, pp.347, 18th Annual IEEE Conference on Computational Complexity (CCC'03), 2003