A CNF formula F is linear if any distinct clauses in F contain at most one common variable. A CNF formula F is exact linear if any distinct clauses in F contain exactly one common variable. All exact linear formulas are satisfiable [4], and for the class LCNF of linear formulas, the decision problem LSAT remains NP-complete. For the subclasses LCNF of LCNF≥k, in which formulas have only clauses of length at least k, the decision problem LSAT≥k remains NP-complete if there exists an unsatisfiable formula in LCNF≥k [3,5]. Therefore, the NP-completeness of SAT for LCNF≥k (k ≥ 3) is the question whether there exists an unsatisfiable formula in LCNF≥k. In [3,5], it is shown that both LCNF≥3 and LCNF≥4 contain unsatisfiable formulas by the constructions of hypergraphs and latin squares. It leaves the open question whether for each k ≥ 5 there is an unsatisfiable formula in LCNF≥k. In this paper, we present a simple and general method to construct minimal unsatisfiable formulas in k-LCNF for each k ≥ 3.
[1]
Stefan Porschen,et al.
On Linear CNF Formulas
,
2006,
SAT.
[2]
Stefan Porschen,et al.
Linear CNF formulas and satisfiability
,
2009,
Discret. Appl. Math..
[3]
Stefan Szeider,et al.
Polynomial-time recognition of minimal unsatisfiable formulas with fixed clause-variable difference
,
2002,
Theor. Comput. Sci..
[4]
Armin Biere,et al.
Theory and Applications of Satisfiability Testing - SAT 2006, 9th International Conference, Seattle, WA, USA, August 12-15, 2006, Proceedings
,
2006,
SAT.
[5]
Hans Kleine Büning,et al.
An efficient algorithm for the minimal unsatisfiability problem for a subclass of CNF
,
1998,
Annals of Mathematics and Artificial Intelligence.