Extending the Known Region of Nonlocal Boxes that Collapse Communication Complexity
Pierre Botteron, Anne Broadbent, Marc-Olivier Proulx
TL;DR
The paper investigates non-signalling boxes (NS) and the phenomenon of communication complexity (CC) collapse, seeking conditions under which CC becomes trivial for any computable function $f$. It extends the BBLMTU protocol by introducing local uniformity and a hierarchy of protocols $\mathcal{P}_k$, enabling distributed computation of $f$ with increasing bias using multiple copies of a box and shared randomness. A key analytic contribution is the bias map $\mu_{k+1}=F(\mu_k)$ with $F(\mu)=\frac{\mu}{16}(A+B-\mu^2(A-B))$, where $A=(\eta_{00}+\eta_{01}+\eta_{10}+\eta_{11})^2$ and $B=2\eta_{00}^2+4\eta_{01}\eta_{10}+2\eta_{11}^2$, derived from the box biases $\eta_{xy}$. The condition $A+B>16$ guarantees three fixed points $\{0,\pm\mu_*\}$ with $0$ repulsive, ensuring convergence to a positive fixed point and proving the existence of a universal $p>1/2$ such that CC collapses for any $f$. Applying the result to two NS-slices, Case 1 (PR–PR′–I) and Case 2 (PR–SR–I), yields explicit analytic boundary inequalities in CHSH coordinates that extend previously known collapsing regions, though such boxes are deemed unlikely to occur in Nature.
Abstract
Non-signalling boxes (NS) are theoretical resources defined by the principle of no-faster-than-light communication. They generalize quantum correlations, and some of them are known to collapse communication complexity (CC). However, this collapse is strongly believed to be unachievable in Nature, so its study provides intuition on which theories are unrealistic. In the present letter, we find a better sufficient condition for a nonlocal box to collapse CC, thus extending the known collapsing region. In some slices of NS, we show this condition coincides with an area outside of an ellipse.
