Superactivation of genuine multipartite Bell nonlocality from two-party entanglement

Markus Miethlinger, Riccardo Castellano, Pavel Sekatski, Nicolas Brunner · 2026

Read Superactivation of genuine multipartite Bell nonlocality from two-party entanglement by Markus Miethlinger, Riccardo Castellano, Pavel Sekatski, Nicolas Brunner (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-17783.txt.

Introduction

Miethlinger, Castellano, Sekatski, and Brunner study when full multipartite Bell nonlocality can emerge from weak parts. Their start states hold entanglement only between pairs, the least one could ask for. Across many copies, that thin base still yields genuine multipartite nonlocality, they prove. They give a workable test to certify the effect on network states. A clean result on repeating the Khot-Vishnoi game backs the proof.

Catalog record

CREATORMarkus Miethlinger, Riccardo Castellano, Pavel Sekatski, Nicolas Brunner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41851
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.17783

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Superactivation of genuine multipartite Bell nonlocality from two-party entanglement Markus Miethlinger, Riccardo Castellano, Pavel Sekatski, Nicolas Brunner Abstract Characterizing the relation between entanglement and Bell nonlocality is a long-standing open problem, notably challenging in the multipartite case. Here, we investigate the effect of superactivation of genuine multipartite nonlocality. Specifically, we show that starting from multipartite states that feature only two-party entanglement (hence almost fully separable), it is possible to obtain GMNL in the many-copy regime. This represents the weakest possible resource for GMNL superactivation. On the technical side, we develop an efficient and practical criterion for certifying GMNL superactivation based on network entangled states, as well as a perfect parallel repetition result for the Khot-Vishnoi Bell game, which are of independent interest. I Introduction Quantum entanglement and Bell nonlocality have long been central to quantum research, from a foundational perspective but also for applications in quantum information processing. While entanglement arises at the level of the mathematical structure of quantum theory, Bell nonlocality is a property of measurement statistics, directly observable in an experiment. Understanding the relation between entanglement and Bell nonlocality is one of oldest open problems in quantum information theory, which is still the focus of current research, see, e.g., 13 ; 3 . It was first discovered by Werner 46 , and later generally proven by Barrett 6 , that certain entangled states—termed “local”—admit a local hidden variable model, meaning that their statistics resulting from arbitrary local measurements can be exactly reproduced via a purely classical model. However, it was shown in turn that Bell nonlocality could nevertheless be activated for certain local entangled states by considering more elaborate Bell tests, involving, e.g., local filtering 35 ; 20 ; 22 , local broadcasting 12 ; 9 ; 40 , catalysis 7 , or in network scenarios 16 . Perhaps the most striking instance of this effect is the superactivation of Bell nonlocality in the many-copy scenario, discovered by Palazuelos 34 . Starting from certain local entangled states \rho , it is possible to obtain Bell nonlocality by performing local (but joint) measurements on the state \rho^{\otimes k} , where k denotes the number of copies. Notably, this is possible for any entangled state with strong enough entanglement, as quantified by the entanglement fraction 15 . Other examples of many-copy activation have also been reported 29 ; 31 ; 36 . In the present work, we address these questions in the multipartite case, where both entanglement and nonlocality display a rich structure 4 and a complex relation 42 ; 2 ; 11 . Our focus is on the connection between entanglement and nonlocality in the many-copy scenario. In particular, we ask what are the minimal resources, in terms of entanglement, in order to obtain the strongest form of Bell nonlocality, namely genuine multipartite nonlocality (GMNL) 41 ? Previous works have shown that GMNL can be superactivated for strongly entangled multipartite states. First, it has been shown that any pure genuine multipartite entangled (GME) state becomes GMNL in the many-copy regime (while this is still open in the single-copy regime) 17 . Second, it is proven that certain mixed states that are GME but not GMNL (as they admit a biseparable local model) become GMNL for a sufficient number of copies 18 . In parallel, it has been shown that biseparable (i.e. non- GME) states become GME in the many-copy regime 48 ; 32 ; 45 . Here we show that GMNL can in fact be superactivated starting from much weaker resources, namely states that feature only two-party entanglement. More precisely, we construct N -partite entangled states \rho that are (N-1) separable, i.e., almost fully separable. Then we show that \rho^{\otimes k} becomes GMNL for some large enough k . Hence, we have identified the weakest possible entanglement resource for superactivating GMNL. On the more technical level, our proof relies on new results on multipartite extensions of bipartite Bell games, and a perfect parallel repetition result for the Khot-Vishnoi (KV) game 28 ; 14 , which are of independent interest. II Setting the scene Consider a set of N distant parties sharing a multipartite entangled state \rho . Each party \mathcal{A}_{i} , upon receiving a classical input x_{i} , performs a measurement, represented by a POVM \{M_{a_{i}|x_{i}}\}_{a_{i}} , and produces a classical output a_{i} . The resulting probability distribution is given by \displaystyle P(\bm{a}|\bm{x})=\Tr(M_{a_1|x_1}\otimes\ldots\otimes M_{a_N|x_N} \rho), (1) where \bm{a}\coloneqq\bigtimes_{i}a_{i} and \bm{x}\coloneqq\bigtimes_{i}x_{i}, . Such a distribution can exhibit different forms of multipartite Bell nonlocality, the strongest one being GMNL. Essentially, one asks here whether the observed distribution could be explained by a classical model, where the parties are merged into two subgroups such that all parties in each subgroup can jointly produce their outcomes based on all inputs. Additionally, probabilistic mixtures over different subgroups are also taken into account. Such “biseparable” local models take the following form \displaystyle\begin{split}&P(\bm{a}|\bm{x})=\sum_{\lambda}\mu(\lambda)P_{bp}(\bm{a}_{\lambda},\overline{\bm{a}_{\lambda}}|\bm{x}_{\lambda},\overline{\bm{x}_{\lambda}},\lambda),\end{split} (2) \displaystyle\begin{split}&P_{bp}(\bm{a}_{\lambda},\overline{\bm{a}_{\lambda}}|\bm{x}_{\lambda},\overline{\bm{x}_{\lambda}},\lambda)=P_{\mathcal{G}(\lambda)}(\bm{a}_{\lambda}|\bm{x}_{\lambda},\lambda)\\ &\hskip 94.0pt\cdot\hskip 2.0pt\hskip 1.0ptP_{\overline{\mathcal{G}(\lambda)}}(\overline{\bm{a}_{\lambda}}|\overline{\bm{x}_{\lambda}},\lambda),\end{split} (3) where we refer to P_{bp}(\bm{a}_{\lambda},\overline{\bm{a}_{\lambda}}|\bm{x}_{\lambda},\overline{\bm{x}_{\lambda}},\lambda) as biproduct behaviors, over the bipartition \mathcal{G}(\lambda)|\overline{\mathcal{G}(\lambda)} , and \bm{x}=\bm{x}_{\lambda}\times\overline{\bm{x}_{\lambda}} , \bm{a}=\bm{a}_{\lambda}\times\overline{\bm{a}_{\lambda}} are the inputs and outputs of all parties, respectively, and \bm{a}_{\lambda},\overline{\bm{a}_{\lambda}} and \bm{x}_{\lambda},\overline{\bm{x}_{\lambda}} collect the inputs and outputs for \mathcal{G}(\lambda),\overline{\mathcal{G}(\lambda)} , respectively. A distribution for which a decomposition of the above form does not exist exhibits GMNL. While several definitions have been proposed 4 , depending on the constraints imposed on the distributions P_{\mathcal{G}(\lambda)}(\bm{a}_{\lambda}|\bm{x}_{\lambda},\lambda) in Eq. ( 2 ), we consider here that no special restrictions are imposed on these distributions, as in the original approach of Ref. 41 . This leads to the strongest notion of GMNL. Such distributions can be obtained via strongly entangled states, such as GHZ states 30 and Dicke states 44 . In particular, the observation of GMNL implies that the underlying state \rho features GME 25 , see, e.g., 27 . The problem of interest in the present work is in fact the inverse question. What are the minimal resources, in terms of entanglement, that can give rise to GMNL? In particular, we want to address this question in the multi-copy Bell scenario, i.e., where the underlying state consists of a finite number of copies of a given resource state \sigma . This means that in Eq. ( 1 ), we now consider states of the form \rho=\sigma^{\otimes k} , where k denotes the number of copies of the resource state \sigma . In such a scenario, each party performs a joint quantum measurement on k local subsystems. As these measurements can be performed in some non-separable basis of the k subsystems, this can lead to the phenomenon of superactivation of Bell nonlocality, first investigated in the bipartite scenario 34 ; 15 and more recently in the multipartite scenario 17 ; 18 . Loosely speaking, this means that starting from a resource state \sigma which is weakly entangled, it is possible to obtain strong nonlocality in the multi-copy regime. Below we present a construction for identifying the minimal resources for superactivation of GMNL. Specifically, we prove that starting from a resource state that has only two-party entanglement, hence is almost fully separable—formally (N-1) -separable 26 —it is possible to superactivate GMNL. This shows that, in the many-copy regime, the strongest form of multipartite Bell nonlocality, i.e., GMNL, can be obtained starting from

Read the complete text