#P is Sandwiched by One and Two #2DNF Calls: Is Subtraction Stronger Than We Thought?

TL;DR

This paper demonstrates that two calls to a restricted #2DNF oracle can capture the class gapP, revealing that subtraction is a powerful operation in counting complexity and leading to new algorithmic and structural insights.

Contribution

It shows that two #2DNF calls suffice to capture gapP via subtraction, and that a single subtraction can replace negation, with implications for sparsification and refined Toda's Theorem variants.

Findings

Two #2DNF calls capture gapP via subtraction.

A single subtraction suffices to simulate negation in #2DNF.

Linear-time reduction preserves sparsity and structural properties.

Abstract

The canonical class in the realm of counting complexity is #P. It is well known that the problem of counting the models of a propositional formula in disjunctive normal form (#DNF) is complete for #P under Turing reductions. On the other hand, #DNF $\in$ spanL and spanL $\not\subseteq$ #P unless NL = NP. Hence, the class of functions logspace-reducible to #DNF is a strict subset of #P under plausible complexity-theoretic assumptions. By contrast, we show that two calls to a (restricted) #2DNF oracle suffice to capture gapP, namely, that the logspace many-one closure of the subtraction between the results of two #2DNF calls is gapP. Because #P $\not\subseteq$ gapP, #P is strictly contained between one and two #2DNF oracle calls. Surprisingly, the propositional formulas needed in both calls are linear-time computable, and the reduction preserves interesting structural as well as symmetry properties, leading to algorithmic applications. We show that a single subtraction suffices to compensate for the absence of negation while still capturing gapP, i.e., our results carry over to the monotone fragments of #2SAT and #2DNF. Since our reduction is linear-time, it preserves sparsity and, as a consequence we obtain a sparsification lemma for both #2SAT and #2DNF. This has only been known for kSAT with k $\geq$ 3 and respective counting versions. We further show that both #2DNF calls can be combined into a single call if we allow a little postprocessing (computable by AC0- or TC0-circuits). Consequently, we derive refined versions of Toda's Theorem: PH $\subseteq$ [#MON2SAT]$^{log}_{TC0}$ = [#MON2DNF]$^{log}_{TC0}$ and PH $\subseteq$ [#IMPL2SAT]$^{log}_{AC0}$. Our route to these results is via structure-aware reductions that preserve parameters like treewidth up to an additive overhead. The absence of multiplicative overhead indeed yields parameterized SETH-tight lower bounds.