Experimental demonstration of full network nonlocality in the bilocal scenario

Emil Håkansson,1, 2 Amélie Piveteau,1 Sadiq Muhammad,1 and Mohamed Bourennane1

1
Department of Physics, Stockholm University, S-10691 Stockholm, Sweden
2
Hitachi Energy Research, Forskargränd 7, 72219 Västerås, Sweden
(Dated: January 23, 2022)

Quantum correlation between nodes in a network which

consist of several independent sources of entanglement
and in multipartite entanglement systems are important
for general understanding of the nature of nonlocality,
quantum information processing and communication. In
previous years, demonstrations of network nonlocality
for bilocal scenarios have been in the focus. Yet,
it has been found that the seminal protocols do not
certify entanglement between end nodes, which otherwise
require all sources in the network to be of nonlocal
(a)
nature. This has motivated for development of even
stronger concept, called full network nonlocality. Here,
we experimentally demonstrate full network nonlocal
correlations in a network. Specifically, we use two pairs of
polarization entangled qubits created by two separate and
independent entanglement sources and a partial Bell state
measurement. Our results may pave the way for the use of
nonlocality test in quantum communication protocols with
a full network nonlocality certification.
Quantum entanglement is one of the fundamental
properties in nature, which leads to nonlocal correlations
between distance objects and has no intuitive explanation (b)
in classical physics. In the wake of Bell’s theorem [1],
which shows incompatibility between the local hidden
variable (LHV) model and the predictions stipulated by
quantum mechanics, there has been a stream of experiments
falsifying LHV models by violating the Bell inequality.
However, it is only recently we have witness loophole-free
confirmation of Bell nonlocality [2], showing the need of
further development of quantum technologies to realize
quantum entangled-based applications. Entanglement across
multiple nodes [3–5] promise to enable a wide range of
applications like secure communication, distributed quantum (c)
computing, cryptography, quantum key distribution, random
number generation and quantum sensing [6–10]. Connection FIG. 1: Bilocal quantum network. Nodes, blue rectangles where A,
of multiple entangled systems into a network structure B, and C indicated Alice ’s , Bob’s, and Charlie’s nodes
is a backbone of future quantum internet [11] which, by respectively. (x, a), (y, b) , and (z, c) are the input and output of
using entanglement-swapping mechanism [12], can in Alice , Bob, and charlie respectively. The parties are connected by
principle enable long-distance quantum key generation for red arrows which symbolize a sharing of a common source. (a)
secure communication [13, 14]. To realize the potential LHV model and its event-ready Bell test, where the region of
influence is steered by one hidden variable λ. (b) BLHV model with
of quantum internet, it is important to establish reliable
two different local hidden variables (λ1 and λ2 ) represented as two
certification methods of non-local correlations between different regions of influence (different shadings). (c) FNN
nodes in the network. Historically, entanglement swapping scenario. The combination with one local hidden variable and an
protocols based on Event-ready Bell test [12] led to the non-local source (N S) (represented by two different shadings)
first bipartite quantum network where conditional successful given rise to local full network model.
outcome from an intermediate node together with correlation
outcome between terminal nodes is used to disprove an independent hidden variables [15]. Recent experiments have
LHV model (see Fig. 1a). Naturally, a general quantum considered this scenario to extend LHV model into bilocal
network consists of truly independent sources, hence it is hidden variable model (BLHV) proposed by [16], further
important to investigate a different scenario based on several investigated by [17–21] and experimentally verified by
2

[22–25] (see Fig.1b). However, bilocal inequality violation network [23]. In that case we need to embrace an even
which disprove BLHV model is not device independent stronger concept in notion of network nonlocality where all
certification of nonlocal sources between terminal nodes [26]. sources are to be of nonlocal nature. Introduced by [26] we
It is therefore of principal importance to find a general model define full network nonlocality (FNN) if the network cannot
which describes a fully non-local network enable to witness be described by allowing at least one source in the network
nonlocal resources between the end nodes. A solution has to be of a local nature, while the rest are characterized to be
been recently proposed by A. Pozas-Kerstjens et al. [26] independent nonlocal (see Fig. 1c). Formalizing this concept
which addresses the shortcoming of BLHV model and lead to bilocal case scenario, the non-FNN correlations are given
into a quantum protocol that can be implemented in photonic by
system for bilocal case scenario. The aim of our work is Z
to experimentally investigate this novel Bell-type inequality p(a, b, c|x, y, z) = dλµ(λ)p(a|x, λ)p(b, c|y, z, λ). (3)
tailored for certifying full network non-locality (FNN). Our
experiment is implemented by using two pairs of polarized
By considering swapping of the sources, it has been
entangled photons in a network configuration with three
shown that produced conditional distributions, given by
nodes, referred shortly as a bilocal network scenario. Here,
p(a, b, c|x, y, z), leads to following FNN witnesses where
two distant parties Alice and Charlie establish a quantum
both need to be violated simultaneously [26]:
communication channel by independently generating
entangled photons. An intermediate party, Bob, performs RC−N S = 2 ⟨A0 B1 C0 ⟩ − 2 ⟨A0 B1 C1 ⟩ + 2 ⟨A1 B0 C0 ⟩ + ⟨A1 B0 C1 ⟩
Bell-state measurement (BSM) using one photon from each
entangled pair. The correlations which are generated between − ⟨B0 ⟩ + ⟨C1 ⟩ [⟨A1 B0 ⟩ + ⟨B0 C0 ⟩ − ⟨C0 ⟩] ≤ 3
all parties lead to violation of the Bell-type inequality, (4)
allowing us to disprove any presence of LHV sources in the
and
network.
RN S−C = 2 ⟨A0 B1 C0 ⟩ − 2 ⟨A0 B1 C1 ⟩ + ⟨A1 B0 C0 ⟩
In comparison to the standard Bell scenario, new forms
+ 2 ⟨A1 B0 C1 ⟩ − ⟨B0 ⟩ + ⟨A1 ⟩ ⟨A1 B0 ⟩ + ⟨A1 ⟩ ⟨B0 C1 ⟩
of nonlocality arising from different quantum network
structures [16, 27–29] are interesting to investigate due to + ⟨A1 ⟩ ⟨C0 ⟩ − ⟨A1 ⟩ ⟨C1 ⟩ − ⟨A1 ⟩ ⟨A1 ⟩ ≤ 3, (5)
their importance for future communication and certification
protocols. Generally, for a bilocal network composed of two where the expectation values are computed in accordance to
sources and two communication parties, each of whom selects [17]
a private input xk and produces an output ak , the correlations X
follow a network local model if they can describe by each ⟨Ax B0 Cz ⟩ = −(1)a+c [p(a, 0, c|x, z)
a,c
source independently emitting a local hidden variables λj
[26]: +p(a, 1, c|x, z) − p(a, 2, c|x, z)]
Z Z
and
p(ā|x̄) = dλ1 µ1 (λ1 ) dλ2 µ2 (λ2 )
X
⟨Ax B1 Cz ⟩ = −(1)a+c [p(a, 0, c|x, z) − p(a, 1, c|x, z)] .
× p(a1 |x1 , λ¯1 )p(a2 |x2 , λ̄2 ), (1)
a,c

where ā = {a1 , a2 } and x̄ = {x1 , x2 , } are sets of output

In our experiment the two entanglement sources emit
and inputs receptively, µ1 , µ2 are probability density functions
singlet state (ψ − ). Bob performs a partial BSM, with three
and λ̄k is a set of local variables associated with the source
different outcomes that are parameterized by ternary bit
that connect to node k. Independence of λj emerges naturally
{0, 1, 2} ∈ b, corresponding to state ϕ+ , ϕ− , and the third
as a construction of independent source in a network. In its
corresponding to undistinguishing between the two remaining
simplest form, with a single source (λ1 = λ2 ), the model
Bell states (ψ + and ψ − ). Alice and Charlie have to choose
reduces to original Bell’s notion of LHV [1, 30]. Extending to
between two possible dichotomic measurements, {x, z} ∈
two independent sources, we can form a three node network
{0, 1}, corresponding to observables A0 = σx , A1 = σz ,
described by two independent hidden variables λ1 and λ2 ,
C0 = σz√+σ x
and C1 = σz√−σ x
and are producing binary
thus form a BLHV model [16]: 2 2
Z outputs {a, c} ∈ {0, 1}. The resulting distribution
√ leads to
p(a, b, c|x, y, z) = dλ1 dλ2 µ1 (λ1 )µ2 (λ2 ) the violation of RC−N S = RN S−C = 5/ 2 ≈ 3.5355
The schematics of our setup is presented in Fig.2. A
× p(a|x, λ1 )p(b|y, λ1 λ2 )p(c|z, λ2 ). (2) femtosecond pulsed laser with repetition rate of 80 MHz
producing ultraviolet light (390 nm) which is split and focused
The bilocal inequalities obtained from BLHV model (2) are into two different setups acting as our two independent
proof of network nonbilocality as it rules out any bilocal sources of entanglement. The first source consists of a
model. Although it is conceptually novel way to violate crossed configuration of two 2 mm thick beta barium borate
Bell-like inequality, simple counterexample shows that it can (BBO) nonlinear crystals producing polarization-entangled
not certify entanglement between the end nodes of a bilocal photon pairs through the process of second order degeneration
 3

Experimental setup for testing FNN model in a bilocal network configuration: The stations of Alice, Bob, and Charlie are highlighted
in blue shading and the two entanglement sources S1 and S2, which connect them, are highlighted in orange. Both sources are pumped by
ultraviolet light centered at a wavelength of 390 nm with a repetition rate of 80 MHz. Source S1 consist of one BBO crystal and source S2
consist of two BBO crystals in cross-configuration. The spatial, spectral and temporal distinguishability between the down-converted photons
is carefully minimized by using narrow band filters, quartz-wedges, and fine alignment translation stages (the last two not included in the
picture). To actively remove the coherence between laser pumping into the two sources we implement a quantum random number generator
(QRNG) connected to a variable phase shifter (see further in SM). Alice and Charlie implement their measurements using half–wave plates,
polarizing beam splitters (PBS) and couplers to single photon detectors (actively quenched Si-avalanche photodiodes, DET) via single mode
fibers. Bob implements his partial BSM, which has two input ports, using a single PBS and two half-wave plates. By using fiber 50:50 beam
splitter (BS) to split each of the four output ports, Bob is able to do partial photon-number resolving detection and therefore gain information
about the two unresolvable Bell states. The observable four-photon coincidence events, which are generated by one simultaneous click for
Alice, Charlie and two clicks for Bob, are registered by field programmable gate array and further the data is processed to calculate conditional
probabilities p(a, b, c|x, y, z). Bob’s outcomes; b = 0 corresponds to projection onto the Bell state |ϕ+ ⟩, b = 1 corresponds to |ϕ− ⟩ and b = 2
corresponds to unresolvable |ψ − ⟩ or |ψ + ⟩.

type-1 spontaneous parametric down conversion (SPDC). The BSM analyzer we can detect, with 50% probability the
second source consists of one 2 mm thick type-2 BBO crystal presence of other two Bell states, ψ ± , although without
generating non-collinear polarization-entangled photons at possibility to distinguish them apart [31]. The photons were
780 nm. To further insure independency between the sources detected with twelve avalanche photodiodes (APD) with an
we introduce decoherence by a movable glass plate in one average efficiency of approximately 55-60 % and the output
of the pumping pulsed laser paths. The glass plate is steered signals from these detectors sent to a field programmable
by a moving motor which receive rotation coordinates from gate array (FPGA) based coincidence counting unit. The
a quantum random number generator, described further total experiment lasted for approximately 15 hours at low
in supplementary material. One photon from each pair is pumping power to avoid creation of multiple photons
sent to a BSM analyzer, where they are interfered. To get pairs. The produced correlations from the experiment
a high-quality interference it is necessary for the photons rendered FNN witnesses of RC−N S = 3.17 ± 0.05 and
from both paths to be indistinguishable in their temporal, RN S−C = 3.17 ± 0.05, more than three standard deviations
spatial, and spectral modes. To increase spectral overlap above FNN inequality (4) and (5) (see supplementary material
we use 3 nm narrowband interference filters (IF). We verify for further details).
indistinguishability by checking four-photon coincidence in
a Hong-Ou-Mandel (HOM) experiment (see supplementary In summary, we have for the first time demonstrated an
material). The measured HOM dip gave us a visibility value experiment to show FNN for bilocal scenario. The resulted
of v = (89 ± 1.3)%. We have measured for each individual violations of RN S−C = 3.17 > 3 and RC−N S = 3.17 > 3
source the visibility in diagonal polarization bases which are well above the classical bound. However, obtained values
resulted in visibility of (99.1 ± 0.1)% and (98.0 ± 0.1)% are lower than predictions RN S−C = RC−N S ≈ 3.5355
for the sources 1 and 2 respectively. It is well know that [26] can be explained by an imperfect HOM-dip visibility.
with linear optics one cannot distinguish between all four Similarly to any Bell-type inequalities, our experiment of
Bell states, leading us to distinguish only two of them, FNN is subject to detection and locality loopholes. From
ϕ± . However, by using beam splitters at the outputs of a more practical perspective, our findings open up for
4

considering complex quantum networks and its connection filtering (10nm at Alice and 3 nm at Bob’s and Charlie’s) the
to FNN. This involves configurations in a hybrid variant photons were collected by three different stations using single
where more than two nodes are connected by nonlocal source mode fibers coupled to avalanche photodetectors. At Alice’s
together with one or several distance nonlocal sources. station, the observable A0 = σx corresponds to half-wave
This will arguably be of great interest from the quantum plate rotated by θ0A = 22.5◦ and A1 = σz corresponds to
information point of view as it also provides a certification half-wave plate rotated by θ1A = 0◦ . At Charlie’s measuring
method to guarantee entanglement between two distance station C0 = σz√+σ 2
x
corresponds to half-wave plate rotated
parties just sharing a BSM device. by θ0 = 11.25 and C1 = σz√−σ
C ◦ x
corresponds to half-
2
C ◦
wave plate rotated by θ0 = −11.25 . Bob’s station has
METHODS only one input, performing a partial BSM which compose of a
polarisation beam splitter together with two half-wave plates
at 22.5◦ . Each measuring configuration takes about 3.5 hours
Experimental details. Correlated photons pair were
to perform and we use fair-sampling assumption.
generated by two separated and independent SPDC sources.
The first source composed of two 2 mm thick BBO nonlinear
crystals in a configuration where one crystal was rotated 90
degrees relative to the other. The second source composed ACKNOWLEDGMENTS
of a single 2 mm thick BBO crystal producing non-collinear
entangled photons. The pumping fields with wavelength of We like to thank Dr. Armin Tavakoli for useful discussions.
λ = 780 nm were adapted in power to generate approximately This work was supported by the Knut and Alice Wallenberg
equal amount of correlated photons from each source. After Foundation through the Wallenberg Center for Quantum
spatial and temporal walkoff compensation and narrowband Technology (WACQT) and the Swedish research council.

[1] J. S. Bell, Physics 1, 195 (1964). [18] A. Tavakoli, P. Skrzypczyk, D. Cavalcanti, and A. Acín, Phys.
[2] B. Hensen, H. Bernien, A. E. Dréau, A. Reiserer, N. Kalb, Rev. A 90, 062109 (2014).
M. S. Blok, J. Ruitenberg, R. F. L. Vermeulen, R. N. [19] N. Gisin, Q. Mei, A. Tavakoli, M.-O. Renou, and N. Brunner,
Schouten, C. Abellán, W. Amaya, V. Pruneri, M. W. Mitchell, arXiv:1702.00333 (2017).
M. Markham, D. J. Twitchen, D. Elkouss, S. Wehner, T. H. [20] F. Andreoli, G. Carvacho, L. Santodonato, R. Chaves, and
Taminiau, and R. Hanson, Nature 526, 682 (2015). F. Sciarrino, arXiv:1702.08316 (2017).
[3] H. J. Kimble, Nature 453, 1023 (2008). [21] A. Tavakoli, N. Gisin, and C. Branciard, arXiv:2006.16694
[4] A. Acin, J. Cirac, and M. Lewenstein, Nature Physics 3, 256 (2020).
(2007). [22] G. Carvacho, F. Andreoli, L. Santodonato, M. Bentivegna,
[5] N. Sangouard, C. Simon, H. de Riedmatten, and N. Gisin, Rev. R. Chaves, and F. Sciarrino, Nat. Commun. 8, 14775 (2017).
Mod. Phys. 83, 33 (2011). [23] D. Saunders, A. Bennet, C. Branciard, and G. Pryde, Science
[6] A. S. S. L. Jiang, J. M. Taylor and M. D. Lukin, Phys. Rev. A Advances 3, e1602743 (2017).
76, 062323 (2007). [24] Q. Sun, Y. Jiang, B. Bai, W. Zhang, H. Li, X. Jiang, J. Zhang,
[7] A. Broadbent, J. Fitzsimons, and E. Kashefi, Annual IEEE L. You, X. Chen, Z. Wang, Q. Zhang, J. Fan, and J. Pan, Nat.
Symposium on Foundations of Computer Science , 517 (2009). Photonics 13, 687–691 (2019).
[8] A. Ekert and R. Renner, Nature 507, 443 (2014). [25] D. Poderini, I. Agresti, G. Marchese, E. Polino, T. Giordani,
[9] N. [Link], J. F. Fitzsimons, and S. C. Benjamin, Phys. A. Suprano, M. Valeri, G. Milani, N. Spagnolo, G. Carvacho,
Rev. X 4, 041041 (2014). R. Chaves, and F. Sciarrin, Nat. Commun. 11, 2467 (2020).
[10] P. Kómár, E. M. Kessler, M. Bishof, L. Jiang, A. Sørensen, [26] A. Pozas-Kerstjens, N. Gisin, and A. Tavakoli, arX-
J. Ye, and M. D. Lukin, Nature Physics 10, 582 (2014). ive:2105.09325 (2021).
[11] D. E. S. Wehner and R. Hanson, Science 362, eaam9288 (2018). [27] T. Fritz, New J. Phys. 14, 103001 (2012).
[12] M. Żukowski, A. Zeilinger, M. A. Horne, and A. K. Ekert, Phys. [28] M.-O. Renou, E. Baumer, S. Boreiri, N. Brunnerr, N. Gisin, and
Rev. Lett. 71, 4287 (1993). S. Beigi, Phys. Rev. Lett. 123, 070403 (2019).
[13] A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991). [29] M.-O. Renou, D. Trillo, M. Weilenmann, L. Thinh, A. Tavakoli,
[14] J. Barrett, L. Hardy, and A. Kent, Phys. Rev. Lett. 95, 010503 N. Gisin, A. Acin, and M. Navascues, arXiv:2101.10873
(2005). (2021).
[15] A. Tavakoli, A. Pozas-Kerstjens, M.-X. Luo, and M.-O. Renou, [30] G. Svetlichnys, Phys. Rev. D 35, 3066 (1987).
arXiv:2104.10700v2 (2021). [31] P. Kwiat and H. Weinfurter, Phys. Rev. A 58, R2623 (1988).
[16] C. Branciard, N. Gisin, and S. Pironio, Phys. Rev. Lett. 104,
170401 (2010).
[17] C. Branciard, C. Rosset, N. Gisin, and S. Pironio, Phys. Rev. A
85, 032119 (2012).