To say simply, can PSPACE problems be written as $\Pi_1$ formula? Or how can these problems be written in terms of (first-order) arithmetic hierarchy?

edit:also currently, by what arithmetic hierarchy formula can P=PSPACE be written?

and what would be the consequence of being able to write P=PSPACE as $\Pi_1$ formula?

  • $\begingroup$ What do you mean by $P=PSPACE$? They're two different complexity classes and it's unknown whether they're equal or not. $\endgroup$ – John Moeller Feb 25 '13 at 5:24
  • 1
    $\begingroup$ Are you familiar with descriptive complexity? $\endgroup$ – JSchlather Feb 25 '13 at 5:34
  • $\begingroup$ I am learning it-and that's why I am asking this question. $\endgroup$ – mars Feb 25 '13 at 7:34

You can get a $\Sigma_2$ sentence for $\mathsf{P}=\mathsf{PSPACE}$, using the same technique as in the blog post, only replace $SAT$ by $QBF$. I doubt there is a known sentence of lower complexity.


$PSPACE \subset R \subset \text{co-}RE = \Pi_1$. $PSPACE$ problems are decidable in finite time, since $PSPACE \subset EXP$. Languages in $R$ are decidable by a Turing machine, and $R=RE\cap\text{co-}RE$, and $\text{co-}RE$ is $\Pi_1$. This implies that any language in $PSPACE$ is also in $\Pi_1$, so the basic answer to your question is "yes."

If you're also asking about $P$, then the answer is also "yes" because $P \subseteq PSPACE$.

  • 2
    $\begingroup$ but what about expressing P=PSPACE as first-order formula? like rjlipton.wordpress.com/2009/05/27/arithmetic-hierarchy-and-pnp $\endgroup$ – mars Feb 25 '13 at 7:37
  • $\begingroup$ As the question is worded, this is a complete answer. $\endgroup$ – András Salamon Feb 25 '13 at 9:31
  • $\begingroup$ @mars Ah, ok. I'm not really qualified to answer that. However, if you wanted to know about something from a blog post, you probably should have provided the link in the question. When you had asked this over on math.sx, I asked for clarification on a couple different points. You could have referenced the blog post then too. $\endgroup$ – John Moeller Feb 25 '13 at 17:55

As already answered by sdcvvc, $\mathrm P=\mathrm{PSPACE}$ can be written as a $\Sigma^0_2$-sentence. On the other hand, if it is equivalent to a $\Pi^0_2$-sentence (let alone $\Pi^0_1$), then this fact does not relativize, by the same argument as in https://mathoverflow.net/questions/57348. A combinatorial proof is given in https://cstheory.stackexchange.com/a/16644/6959.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy