0
$\begingroup$

2QBF is the following problem: given a CNF formula $\psi$ on $2n$ variables, determine the truth value of

$$\forall x \in \{0,1\}^n . \exists y \in \{0,1\}^n . \psi(x,y).$$

Question: Is 2QBF in $P^{NP}$ (or, expected to be in $P^{NP}$, under standard conjectures)?

Motivation: If the answer were yes, it would suggest a way to solve 2QBF, using multiple invocations of a SAT solver. Given that SAT solvers are surprisingly effective on some problems, that might be of interest in situations where we need to solve 2QBF.

(2QBF is a version of QBF restricted to two alternating quantifiers. QBF is known to be PSPACE-complete, but 2QBF seems easier than QBF.)

$\endgroup$
1
  • 2
    $\begingroup$ 2QBF is complete for the 2nd level of the polynomial hierarchy ($NP^{NP}$). So, if it is shown to be in ($P^{NP}$), wouldn't the polynomial hierarchy collapse (which is considered unlikely)? $\endgroup$ Commented Feb 22, 2017 at 19:18

2 Answers 2

1
$\begingroup$

Determining the truth of $\exists \forall. \phi$ formulas is $\Sigma_2^P$-complete whereas determining the truth of $\forall \exists. \phi$ formulas are $\Pi_2^P$-complete. Therefore, unless $\Sigma_2^P=P^{NP}$ (resp., $\Pi_2^P=P^{NP}$) neither of these problems can be solved in $P^{NP}$.

$\endgroup$
1
  • $\begingroup$ user109711 has explained to me that $\exists \forall \phi$ formulas are in NP if $\phi$ is in CNF: you can check the validity of $\forall \phi$ by checking separately for each clause if it holds universally. $\endgroup$
    – D.W.
    Commented Jun 20, 2020 at 19:38
1
$\begingroup$

2QBF (in the form presented in the question) is $\Pi_2^P$-complete. Also, $P^{NP} = \Delta_2^P = P^{coNP}$ and $\Pi_2^P = coNP^{coNP}$.

Therefore, if 2QBF is in $P^{NP}$, then $\Pi_2^P = \Delta_2^P$, from which it follows that $\Pi_2^P = \Pi_3^P$ (since $P^{coNP} = coNP^{coNP}$ implies $\Pi_3^P = coNP^{coNP^{coNP}} = coNP^{P^{coNP}} = coNP^{coNP} = \Pi_2^P$), which means that the polynomial hierarchy collapses -- something that is conjectured not to happen.

Therefore, under standard conjectures (that the polynomial hierarchy doesn't collapse), 2QBF is not in $P^{NP}$. In other words, there is no general way to solve 2QBF using polynomially invocations of a SAT solver.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.