7

I read in a paper a proof where you can reduce a $3$-SAT problem into $2$-SAT + HORN-SAT clauses.

$2$-SAT + HORN-SAT is therefore, NP-complete.

$2$-SAT, HORN-SAT, DUAL HORN-SAT, XOR-SAT are all in P.

I would like to know, if there is a Polynomial time algorithm for the conjuntion of $2$-SAT and XOR-SAT formulas.

yugikaiba
  • 176

1 Answers1

6

There is no polynomial-time procedure for deciding the satisfiability of 2-SAT + XOR-SAT unless P = NP. So, probably not. 2-SAT + XOR-SAT is easily proven NP-complete by direct polynomial reduction from 3-SAT.

In a 3-SAT instance, each 3-CNF clause $$(x_1 \lor x_2 \lor x_3)$$ can be rewritten into the equisatisfiable 2-SAT + XOR-SAT expression $$(x_1 \lor \overline{y}) \land (y \oplus x_2 \oplus z) \land (\overline{z} \lor x_3)$$ with $y$ and $z$ as new variables that appear nowhere else in the formula.

Kyle Jones
  • 1,921
  • 1
    Thanks for answering. I was looking for either a 3-sat reduction or an algorithm. – yugikaiba Dec 28 '20 at 03:55
  • 1
    $x_1$=False, $x_2$=False, $x_3$=False and the formulas are not equisatisfiable. – yugikaiba Dec 28 '20 at 04:35
  • 1
    the second expression in CNF is equivalent to: $$(\overline{x_1} \lor \overline{y} \lor z) \land (\overline{x_1} \lor y \lor \overline{z}) \land (x_1 \lor \overline{y} \lor \overline{z}) \land (x_1 \lor y \lor z) \land (x_2 \lor \overline{y}) \land (x_3 \lor \overline{z})$$ which is not equisatisfiable with the first. – yugikaiba Dec 28 '20 at 08:01
  • 1
    @yugikaiba For $x_1 = x_2 = x_3 = \textrm{False}$, neither formula is satisfiable: the second reduces to $\overline y ∧ (y ⊕ z) ∧ \overline z$. This answer is correct. – Anders Kaseorg Dec 31 '20 at 22:16
  • true, thank you – yugikaiba Jan 01 '21 at 11:23