Abstract
Superposition, entanglement and nonlocality constitute fundamental features of quantum physics. The fact that quantum physics does not follow the principle of local causality^{1,2,3} can be experimentally demonstrated in Bell tests^{4} performed on pairs of spatially separated, entangled quantum systems. Although Bell tests, which are widely regarded as a litmus test of quantum physics, have been explored using a broad range of quantum systems over the past 50 years, only relatively recently have experiments free of socalled loopholes^{5} succeeded. Such experiments have been performed with spins in nitrogen–vacancy centres^{6}, optical photons^{7,8,9} and neutral atoms^{10}. Here we demonstrate a loopholefree violation of Bell’s inequality with superconducting circuits, which are a prime contender for realizing quantum computing technology^{11}. To evaluate a Clauser–Horne–Shimony–Holttype Bell inequality^{4}, we deterministically entangle a pair of qubits^{12} and perform fast and highfidelity measurements^{13} along randomly chosen bases on the qubits connected through a cryogenic link^{14} spanning a distance of 30 metres. Evaluating more than 1 million experimental trials, we find an average S value of 2.0747 ± 0.0033, violating Bell’s inequality with a P value smaller than 10^{−108}. Our work demonstrates that nonlocality is a viable new resource in quantum information technology realized with superconducting circuits with potential applications in quantum communication, quantum computing and fundamental physics^{15}.
Similar content being viewed by others
Main
One of the astounding features of quantum physics is that it contradicts our common intuitive understanding of nature following the principle of local causality^{1}. This concept derives from the expectation that the causes of an event are to be found in its neighbourhood (see Supplementary Information section I for a discussion). In 1964, John Stewart Bell proposed an experiment, now known as a Bell test, to empirically demonstrate that theories satisfying the principle of local causality do not describe the properties of a pair of entangled quantum systems^{2,3}.
In a Bell test^{4}, two distinct parties A and B each hold one part of an entangled quantum system, for example, one of two qubits. Each party then chooses one of two possible measurements to perform on their qubit, and records the binary measurement outcome. The parties repeat the process many times to accumulate statistics, and evaluate a Bell inequality^{2,4} using the measurement choices and recorded results. Systems governed by local hidden variable models are expected to obey the inequality whereas quantum systems can violate it. The two underlying assumptions in the derivation of Bell’s inequality are locality, the concept that the measurement outcome at the location of party A cannot depend on information available at the location of party B and vice versa, and measurement independence, the idea that the choice between the two possible measurements is statistically independent from any hidden variables.
A decade after Bell’s proposal, the first pioneering experimental Bell tests were successful^{16,17}. However, these early experiments relied on additional assumptions^{18}, creating loopholes in the conclusions drawn from the experiments. In the following decades, experiments relying on fewer and fewer assumptions were performed^{19,20,21}, until loopholefree Bell inequality violations, which close all major loopholes simultaneously, were demonstrated in 2015 and the following years^{6,7,8,9,10}; see ref. ^{22} for a discussion.
In the development of quantum information science, it became clear that Bell tests relying on a minimum number of assumptions are not only of interest for testing fundamental physics but also serve as a key resource in quantum information processing protocols. Observing a violation of Bell’s inequality indicates that the system possesses nonclassical correlations, and asserts that the potentially unknown quantum state has a certain degree of entanglement and purity. This assessment, based on the observed correlations between the chosen input (the choice of measurement basis) and recorded output values (the measurement outcome) of the test, does not rely on knowledge of the inner workings of the system: a property known as device independence^{23}. This allows identifying quantum states and measurements^{24}, certifying the correct functioning of quantum computing devices^{25}, and establishing common and secret keys between two parties with only limited assumptions about the used devices^{26}. Further applications of Bell tests include deviceindependent randomness generation and expansion, extending a given random bit string in a certified manner^{27,28} and randomness amplification, improving the quality of a source of randomness in a certified manner^{29,30}, which is a task impossible to achieve by purely classical means.
Deploying nonlocality as a new resource in the context of superconducting circuits enables new applications in a system that is well set for creating largescale quantum computers^{11,31} and provides quantum communication capabilities. In addition, nonlocal Bell tests with superconducting circuits are unique as a macroscopic quantum system^{32,33,34} is used, which is controlled, entangled and read out exclusively using microwave frequency radiation rather than optical frequency fields.
With superconducting circuits, Bell tests were performed that closed the fairsampling (or detection) loophole^{35}, supported the assumption of measurement independence with human choices^{36}, and used qubits connected by an onchip 78cmlong transmission line^{37}. Whereas these experiments all relied on additional assumptions, in this work, we set out to demonstrate a loopholefree violation of Bell’s inequality using superconducting circuits. The Methods section provides a very brief introduction to the basic properties of superconducting qubits.
Addressing the locality loophole^{5} (Supplementary Information section I) in a Bell test with superconducting circuits, typically housed in their individual cryogenic systems, is particularly challenging, as it requires to entangle a pair of qubits located at two sites A and B separated by a large physical distance d with high concurrence \({\mathcal{C}}\) of the entangled state, where \({\mathcal{C}}\) (refs. ^{38,39}) is a measure of the degree of entanglement present in the system. An individual trial of a Bell test begins at time t_{⋆} = 0 with the choice of a pair of input bits (a, b), which determine the basis in which the quantum state of each of the two entangled qubits is read out (Fig. 1). To support the assumption of measurement independence, local basis choices are realized using random number generators (RNGs). If the sites A and B are separated from each other by a sufficiently large distance d, the exchange of information between A and B, occurring at most at the speed of light c, is prohibited for times t < t_{d} = d/c according to the laws of special relativity. If the measurement outcomes are obtained during this time interval, the spatial separation thus ensures that the chosen measurement bases and the corresponding measurement outcomes by the party at one site are unknown to the party at the other site, thereby closing the locality loophole.
For each trial of the Bell test, a highfidelity measurement of the quantum state of the qubits at A and B, which is designed to terminate at time t < t_{d}, is performed. The readout of the qubits results in outcomes x and y taking on values of +1 or −1 if the qubit is detected in the ground \(\leftg\right\rangle \) or excited state \(\lefte\right\rangle \), respectively. Including each and every measurement outcome in the analysis of the Bell test closes the fairsampling loophole^{40,41} (Supplementary Information section I). Furthermore, the memory loophole is closed by statistically analysing the input and output data without assuming that individual trials of the Bell test are independent and identically distributed^{18}.
To evaluate the result of a Bell test performed in this manner, the averages of the product of the individual measurement outcomes ⟨xy⟩_{(a,}_{b)} at sites A and B are calculated to determine the Clauser–Horne–Shimony–Holt (CHSH) value^{4} S = ⟨xy⟩_{(0,}_{0)} − ⟨xy⟩_{(0,}_{1)} + ⟨xy⟩_{(1,}_{0)} + ⟨xy⟩_{(1,}_{1)} given the four possible combinations of measurement basis choices (a, b). If the properties of the system were described by a local hidden variable model^{2}, one would find ∣S∣ ≤ 2, whereas any value larger than two indicates a violation of Bell’s inequality. The maximum value of ∣S∣ allowed by quantum physics is \(2\sqrt{2}\).
In the following, we discuss how we fulfil the requirements outlined here for realizing a Bell test with superconducting circuits closing the locality, fairsampling and memory loopholes and supporting measurement independence all at the same time.
Requirements
In a Bell test using an entangled pair of qubits, the degree to which Bell’s inequality can be violated depends on the concurrence \({\mathcal{C}}\) of the entangled state and the individual qubitreadout fidelity \({{\mathcal{F}}}_{{\rm{r}}}^{({\rm{A}},{\rm{B}})}\). Together, these quantities constrain the maximally achievable Bell parameter to^{42}
Thus, the CHSH inequality can only be violated if the average readout fidelity \({{\mathcal{F}}}_{{\rm{r}}}=\sqrt{{{\mathcal{F}}}_{{\rm{r}}}^{{\rm{A}}}{{\mathcal{F}}}_{{\rm{r}}}^{{\rm{B}}}}\) exceeds roughly 84% and the concurrence \({\mathcal{C}}\) exceeds roughly 0.7, so that \({S}^{\max } > 2\), as shown in the contour plot in Fig. 2a.
Addressing the above requirement, previous experiments achieved remote entanglement of superconducting qubits with sufficiently large concurrence in a single dilution refrigerator^{12,37,43,44} and in two refrigerators connected across a distance of 5 m using a cryogenic link^{14}. In the experiments we present here, we create entanglement over much larger linear distances. In addition, singleshot readout of superconducting qubits was demonstrated in an integration time of 50 ns with fidelity \({{\mathcal{F}}}_{{\rm{r}}}\) exceeding 98% (ref. ^{13}). In a Bell test that closes the locality loophole, minimizing the duration of the readout reduces the distance d required between the two parties to provide spacelike separation.
Given the expected readout time of roughly 50 ns, the time required for choosing the measurement bases at random (roughly 29 ns, Supplementary Information section II), and accounting for a margin for signal propagation times, we chose to build a cryogenic system housing superconducting circuits at a linear physical distance of roughly d = 30 m (Fig. 3). This provides a time budget t_{d} in excess of 100 ns for the Bell test.
In our experiments, a pair of dilution refrigerators, one at site A and one at B, each house a superconducting qubit with circuitry for local readout and remote entanglement^{12,13,14} cooled down to about 15 mK. In a unique setup (Fig. 3), we connect the two circuits to each other over a distance of 30 m using a cryogenic quantum microwave channel^{14} realized as a superconducting aluminium waveguide. We cool the waveguide to temperatures of a few tens of millikelvin at which its loss is less than 1 dB per km (refs. ^{14,45}) and its thermal occupation is negligible.
To successfully operate this system, we minimized the heat load at each temperature stage using highreflectance materials combined with superinsulation for radiation shielding. We designed vertical support structures between the individual shielding stages to minimize thermal conductivity while providing mechanical stability. The system withstands thermal contractions by leveraging flexible thermal connections formed by braids and mobile mechanical supports. We maximize heat flow along the link modules by using highconductivity copper and minimizing thermal contact resistances between adjacent link elements. At the midpoint between sites A and B, a pulse tube cooler provides an additional heat sink for the thermal radiation incident on the 50 and 4 K radiation shields. At 30 m in length and with a total mass exceeding 1.3 tons of radiation shields cooled to below 80 K, roughly 90 kg of which are cooled to below 50 mK, this constitutes a largescale cryogenic system operating at millikelvin temperatures^{46}; see Supplementary Information section III for details.
At each of the sites A and B, we operate a transmonstyle qubit whose state and transition frequency is controlled on nanosecond time scales using amplitude and phasemodulated microwave pulses and magneticflux bias pulses. We read out the state of each qubit using a resonator combined with a Purcell filter^{12,13,14}. For the entanglement protocol, we make use of a photontransfer resonator, also combined with a Purcell filter, which we couple using a coaxial line to the aluminium waveguide connecting the two sites^{12,14}. Both qubits and their support circuitry are fabricated on two nominally identical chips (Supplementary Information section IV).
Bell test
In each individual trial of the Bell test experiment, we deterministically generate a Bell state \(\left{\psi }^{+}\right\rangle =\frac{1}{\sqrt{2}}(\leftge\right\rangle +\lefteg\right\rangle )\) between the stationary transmon qubits at sites A and B using direct photon exchange^{12,47} (Methods). Performing quantum state tomography of the states of the qubits at sites A and B, we experimentally achieve a Bell state fidelity of \({{\mathcal{F}}}_{{\rm{s}}}^{\left\,\,{\psi }^{+}\right\rangle }=80.4 \% \), corresponding to a concurrence^{38} of 0.765 (Fig. 2b) when correcting for readout errors. The experiments were performed with two independent phase synchronized setups separated by 30 m (Supplementary Information section V). Without readout error correction we find \({{\mathcal{F}}}_{{\rm{s}}}^{\left\,\,{\psi }^{+}\right\rangle }=78.9 \% \) and \({\mathcal{C}}=0.689\). The concurrence of the entangled state is sufficiently high for violating Bell’s inequality (Fig. 2a) and is on par with previous experiments using the same approach for entangling qubits in a single cryostat^{12} and in two cryostats connected across a distance of 5 m (ref. ^{14}). The infidelity is dominated by photon loss induced by a circulator used to extract photons from the waveguide for characterization of the entanglement protocol in all three experiments^{12,14} (Supplementary Information section VI). As outlined in the Methods section, we rotate each qubit state along the y axis of the Bloch sphere to maximize the Bell violation. After entangling the distant qubits at sites A and B, we are set to perform the timingcritical part of the loopholefree Bell test as indicated in the space–time diagram in Fig. 4.
To generate the input bits a and b for the measurement basis choice we use a RNG at each node^{48}. We consider the start event of each trial of a Bell test as being marked in space and time by the earlier of the two events corresponding to the creation of a random number in each RNG. At node A (B), a random number is generated at the location indicated by a red (blue) star in the space–time diagram in Fig. 4 at a distance of about 2 m from the corresponding qubit housed in its dilution refrigerator. The random number a (b) becomes available as a voltage pulse at the output of the RNG 17.10 ± 0.14 ns after this event (yellow section in Fig. 4). This pulse controls a microwave switch that conditionally passes a microwave basisrotation pulse to the qubit at A (B). We provide additional information on the random basis choice in the Methods section and in Supplementary Information section II.
To achieve a signal propagation delay of the basis choice pulses applied to each qubit of only 14 ns (first turquoise section in Fig. 4), we pass microwave signals roughly along the line of sight connecting the qubits at A and B from the roomtemperature switch through a sideaccess port into the cryogenic system (Supplementary Information section III). The random basis selection pulse (grey) applied to the qubit has a duration of 12 ns.
After the microwave pulse has fully rotated both qubit states into the randomly chosen basis, we read out the qubits at A and B by applying a microwave tone to their dedicated readout resonators. We detect the amplitude and phase of the readout pulse after several stages of amplification (green section in Fig. 4), record it with a digitizer (analoguetodigital converter, ADC) and postprocess the data with a field programmable gate array (FPGA)^{13}. We achieve singleshot readout fidelities of \({{\mathcal{F}}}_{{\rm{r}}}^{{\rm{A}}}=99.05 \% \) and \({{\mathcal{F}}}_{{\rm{r}}}^{{\rm{B}}}=97.60 \% \) in only 50 ns integration time (Supplementary Information section VII).
As done with the random basis choice signals, we route the readout signals through the side ports of the dilution refrigerators at sites A and B. In this way, we minimize to 14 ns the propagation delay towards the ADC and FPGA, located at about 1 m physical distance from the qubits, indicated by a cross in Fig. 4. We consider the measurement result to be fixed at time t_{×}, the moment when the last part of the measurement signal arrives at the input of the ADC to be digitized, see Supplementary Information section VIII B for a discussion of this choice.
In each Bell test experiment, we recorded the basis choices (a, b) and the corresponding readout result (x, y) for all n trials. On the basis of these values, we calculated the averages ⟨xy⟩_{(a,}_{b)} for all four combinations of measurement basis choices taking into account all n trials, thus closing the fairsampling loophole^{40,41}.
In each of four consecutive experiments, we ran \({n}_{\max }={2}^{20}=1,\,048,\,576\) individual trials of a Bell test for a total time of about 20 minutes. In the first experiment, we swept the angle θ between the two randomly chosen measurement bases (Fig. 1) across a full period. Plotting ⟨xy⟩_{(a,}_{b)} for all four input bit combinations versus θ we observe the expected sinusoidal oscillations^{4}, offset from each other by π/2 (Fig. 5a). Ideally, ⟨xy⟩_{(a,}_{b)} oscillates between +1 and −1. The observed reduction in contrast is due to the finite concurrence of the initial entangled state and readout errors. We note that we have calibrated out an experimental phase offset of θ_{0} = 160.0° between the two sites (Supplementary Information section IX). We find good agreement between the experimental data and a master equation simulation (Fig. 5a,b and Supplementary Information section VI).
On the basis of these data, we then calculate the S value as a function of θ and observe its sinusoidal oscillation with a maximum and a minimum of S occurring at \({\theta }_{{S}_{max}}=\,\pi /4\) and \({\theta }_{{S}_{min}}=\pi \pi /4\) (Fig. 5b), offset by π as expected. Here, we evaluate roughly 60,000 trials for each angle. We find that ∣S∣ exceeds two at both values \({\theta }_{{S}_{\max /\min }}\), violating the Bell inequality.
Near \({\theta }_{{S}_{\max /\min }}\) we perform a set of measurements with step size θ = π/32 and determine the S value from roughly 80,000 trials at each value of θ. We observe that several data sets clearly violate the Bell inequality for angles around \({\theta }_{{S}_{\max /\min }}\) (Fig. 5c,d). From the data set taken at the offset angle \({\theta }_{{S}_{max}}\) we find a maximum violation of S = 2.082 ± 0.012 > 2.
In a final experiment performed at the single angle of \({\theta }_{{S}_{max}}\), we acquire Bell test data for \({n}_{\max }\) trials, yielding S = 2.0747 ± 0.0033, which exceeds two by more than 22 standard deviations. In this experiment with superconducting circuits, we reject the null hypothesis corresponding to Bell’s inequality being satisfied with a P value smaller than 10^{−108} (Methods), which is small in comparison to P values reported for Bell tests closing all major loopholes in the literature (Supplementary Information section VIII A). The statistical method used here is robust to memory effects (Supplementary Information section X).
Finally, we verify that the locality loophole is closed by measuring the physical distance d separating the two pairs of points in space marked by stars (Fig. 4) defining the start of the Bell test trial at t = t_{⋆} = 0 from the points in space marked by stars defining the end of the trial at time t_{×}. Using the methods described in the Supplementary Information section XI, we find the shorter of these two distances to be d = 32.824 m ± 4.6 mm yielding a time budget of t_{d} = 109.489 ± 0.015 ns for the Bell test to close the locality loophole. Using independent measurements, we determine the total duration of the Bell test trial to be t_{×} − t_{⋆} = 107.40 ± 0.26 ns < t_{d} (Supplementary Information section XI), therefore closing the locality loophole with a margin of roughly eight standard deviations. These timing margins are similar to those achieved in loopholefree Bell tests reported in the literature (Supplementary Information section VIII B).
To formulate our conclusion, we assume that we can precisely determine the space–time description of the events at hand and that the RNG produced independent, free random bits. Ultimately, such assumptions constraining the conclusion cannot be fully avoided, even in principle^{18}. Under these assumptions, we find that our observation of the violation of Bell’s inequality with superconducting circuits is incompatible with an explanation satisfying the principle of local causality.
Performance and outlook
Previous loopholefree Bell tests using polarizationencoded optical photons as qubits, typically violated Bell’s inequality with a lower margin^{7,8,9} than our experiment (S = 2.0747) whereas atomic and solidstate systems^{6,10} realized higher violations. By reducing the loss in the channel connecting the two qubits in our setup and thus increasing the Bell state fidelity, we expect Bell violations with S > 2.4 to be achievable in future experiments while closing all major loopholes. We plan to reach this target by omitting the circulator from the waveguide, as in ref. ^{37}, and using both a lowloss printed circuit board and superconducting microwave cables connecting the sample mount to the waveguide. With these measures, we estimate to reduce the photon loss by up to a factor of four to about 5%. Alternatively, a heralded entangling method, avoiding loss but effectively reducing the repetition rate, can be implemented for the same purpose^{12,49}. Such improvements may enable protocols requiring larger Bell violations, such as deviceindependent quantum key distribution^{50}, to be executed between superconducting quantum processors connected in a network.
Because the experiment presented here operates at a repetition rate of 12.5 kHz, which is larger than those of loopholefree Bell tests with atomic and other solidstate systems^{6,10}, we achieve highly statistically significant Bell inequality violations in only a few minutes. This is similar to experiments performed with polarizationencoded optical photons that have even higher repetition rates^{7,8,9}. Further improvements in the repetition rate of our experiment seem feasible, up to the inverse of the duration of the pulse sequence used. In Supplementary Information section VIII we compare in detail the performance metrics of published Bell tests that also used a minimal set of assumptions.
To implement deviceindependent quantum information processing protocols, it is desirable to simultaneously achieve high Bell violations and high repetition rates. The setup demonstrated in our experiments provides an interesting combination of those metrics allowing us to visualize implementing a variety of deviceindependent quantum information processing protocols^{26,27,28,29,30} with superconducting circuits, a promising candidate for largescale quantum computing systems^{11,31}.
In addition, our experiment demonstrates that quantum information can be transmitted between superconducting circuits housed in cryogenic systems separated by tens of metres, going beyond our previous work on a metrescale system^{14}. Interconnected cryogenic systems may indicate a pathway towards realizing larger scale quantum computing systems using quantum microwave local area networks^{51}, for example, within a quantum computing centre. The setup also enables the exploration of nonlocal quantum physics with degrees of freedom that couple to microwave photons such as mechanical resonators or spins.
Methods
Superconducting qubits
For completeness, we discuss a few key features of superconducting quantum electronic circuits here. Superconducting qubits are anharmonic quantum oscillators with circuit parameters chosen to realize resonance frequencies in the gigahertz frequency range (ref. ^{52} and references therein). The nonlinearity of the circuit is provided by an ideally lossless Josephson element^{53}, which, in our experiments, is realized as a pair of Josephson tunnel junctions arranged in a superconducting quantum interference device loop shunted by a large capacitor jointly forming a transmon qubit^{54}. The effective Hamiltonian of the circuit is governed by a cosinusoidal potential that hosts a set of bound states, the lowest two of that form the computational basis states of the qubit (labelled \(\leftg\right\rangle \) and \(\lefte\right\rangle \)). The second excited state (labelled \( \,f\rangle \)) can be used, for example, as an auxiliary state to realize twoqubit gates^{55}, or, as in this paper, for emitting photons with a controlled temporal mode profile^{12,47} when coupling the qutrit strongly to a superconducting resonator (for example, ref. ^{56} and references therein). Here, as well as elsewhere in this publication, we refer to the quantum bit as a qutrit when we refer to its lowest three energy eigenstates. To minimize thermal excitation of the qubit and also minimize losses in the superconducting materials used to realize the qubits and the 30mlong waveguide, we operate the devices at temperatures around 15 mK (ref. ^{25}).
Generation of the remote, twoqubit entangled state
We generate the Bell state \(\left{\psi }^{+}\right\rangle =(\leftge\right\rangle +\lefteg\right\rangle )/\sqrt{2}\) between the two remote qubits A and B using a deterministic scheme based on the exchange of a single photon^{57}, as demonstrated with superconducting circuits, for example, in ref. ^{12}. In this protocol, qubit A is first entangled with a propagating, ideally timereversalsymmetric photon in a driven coherent emission process. The propagating photon is then deterministically absorbed in a timereversed process at qubit B creating the desired entangled state.
The pulse sequence used in the process is about 400 ns long and ends at roughly t = 16 ns after the process initiating the random basis choice starts, see Supplementary Information section VI for details. We create the entangled state as an input resource to the presented Bell test experiment. We consider the entanglement process itself to be independent of the timing constraints of the loopholefree Bell test (Methods and Supplementary Information sections II and XI).
We characterize the created Bell state using quantum state tomography (Fig. 2b) for which we rotate the density matrix in postprocessing by an angle θ_{0} around the z axis to maximize the Bell state fidelity. θ_{0} is the experimental offset angle between the two setups A and B (Supplementary Information section IX). We also perform a master equation simulation of the Bell state generation protocol and find good agreement with the experimental data (red wireframes in Fig. 2b), characterized by the small trace distance \(\sqrt{{\rm{Tr}}( \rho {\rho }_{{\rm{sim}}}{ }^{2}\,)}=0.077\).
Optimizing the measurement basis
Photon loss and qubit decay are the dominant mechanisms reducing the fidelity of the experimentally created Bell state. These processes create an asymmetry between the qubit excited and ground states participating in the Bell test. To reduce this detrimental effect and maximize the S value, we perform the Bell test by choosing measurement bases in the xy plane of the Bloch sphere, where each basis is affected equally by photon loss and qubit decay. To do so, we apply a (π/2)_{x} basisrotation pulse to qubit A, and a (π/2)_{x+θ} pulse to qubit B in each experimental trial just after the \(\left{\psi }^{+}\right\rangle \) Bell state preparation and before applying the pulse implementing the random measurement basis choice. Here, θ denotes the angle between the two randomly chosen measurement bases, as introduced in the main text. Effectively, this pulse sequence generates the Bell state \(\left{\varphi }^{+}\right\rangle =(\leftgg\right\rangle +\leftee\right\rangle )/\sqrt{2}\).
Random basis selection and measurement independence
As in previous loopholefree Bell tests^{6,7,8,10}, we use wellcharacterized, fast physical RNGs^{48} to support the measurement independence assumption. The basis choice bits a and b are generated by a dedicated RNG at each node. Each RNG contains eight quantum entropy sources, each composed of a laser to produce phaserandomized pulses, an interferometer, fast linear detection, onebit digitization and a parity calculator. We assume that the extracted random bits are independent from all previous events. Supplementary Information section II describes support for this assumption and in Supplementary Information section X we perform a statistical analysis.
The RNG output bit is encoded as a voltage, and controls a singlepole singlethrow microwave switch, which (with input high) blocks or (with input low) passes a microwave (π/2)_{y} pulse to induce a basis change of the corresponding qubit. We discuss the basis selection further in Supplementary Information section II.
Because every choice of measurement leads to a recorded experimental trial in our experiment, as in refs. ^{7,8}, the RNG properties offer direct support for the validity of the measurement independence condition. This differs from the situation in eventready Bell tests^{6,10}, in which a heralding event—the result of a joint photon measurement at a middle station—determines whether or not a given experimental trial is recorded for analysis. Such selection opens the possibility for the heralding event to select trials as a function of the measurement choices, if these are not spacelike separated from the joint measurement. In this situation, the measurement independence condition can be violated even in presence of perfect RNGs. For this reason, in eventready Bell tests it is important that the heralding event is spacelike separated from the events marking the creation of the randominput bits. Because we use a deterministic entanglement generation protocol that is not dependent on the outcome of any measurement, such considerations related to a heralding event do not play a role in our experiment.
Note that when focusing on specific families of local hidden variable models, measurement independence can sometimes be supported by spacelike separation alone. This is the case for the model introduced in Scheidl et al.^{58}, and relevant for Bell tests with entangledphoton pairs, where the hidden variable λ is assumed to be created at the photon pair generation event independently of any past event. With that assumption, spacelike separation between the pair generation and the measurement choice events guarantees measurement independence. This spacelike separation is achieved also in refs. ^{7,8}. Assuming λ is produced along with the photon pairs, but relaxing the assumption that λ is independent of past events, a photonic Bell test with the same spacelike separation between the pair generation and the settings choice events can exclude local causal models in which the photon pairs influence the measurement choices, but not local causal models in which earlier events influence both the photons and the basis choices. Because such past influences could be arbitrarily far in the past, space–time conditions cannot be used to fully support the assumption of measurement independence beyond the model of Scheidl et al.^{58}, even in photonic Bell tests. For this reason, we do not attempt to create a spacelike separation between the entanglement generation and the basis choices, but rather support the assumption of measurement independence by using wellcharacterized RNGs^{48}, as has been done in previous Bell tests since the pioneering work of Weihs et al.^{20}.
The P value as a statistical metric
Early Bell test experiments typically used the standard deviation as a metric to discuss the statistical significance of an observed Bell inequality violation. This approach, however, comes with two limitations. The first is that when using the notion of standard deviation, we implicitly assume that the underlying measurement data are Gaussian distributed. This assumption is only justified in the limit of infinitely many trials, but in experiments a finite number of trials are executed. A statistical analysis of the Bell test based on standard deviations may therefore overestimate the statistical significance of the result^{5,59}. The second limitation is that the notion of standard deviation relies on the assumption that the result of the kth trial is independent of the basis choices and measurement results of the previous k − 1 trials, which opens up the memory loophole^{18}. These two limitations can be addressed by the statistical analysis of the result through the calculation of a P value according to a method that does not rely on any of the aforementioned assumptions. Therefore, the calculation of P values in the context of Bell tests is now an established practice^{6,7,8,9,10}. In this context, the P value is a metric of the probability with which data as extreme as those observed could have been produced by a local causal model (see Supplementary Information section X for details).
Data availability
All data are available from the corresponding authors upon reasonable request.
References
Einstein, A., Podolsky, B. & Rosen, N. Can quantummechanical description of physical reality be considered complete? Phys. Rev. 48, 696 (1935).
Bell, J. S. On the Einstein Podolsky Rosen paradox. Physics 1, 195 (1964).
Bell, J. S. in Speakable and Unspeakable in Quantum Mechanics: Collected Papers on Quantum Philosophy 2nd edn, Ch. 24, 232–248 (Cambridge Univ. Press, 2004).
Clauser, J. F., Horne, M. A., Shimony, A. & Holt, R. A. Proposed experiment to test local hiddenvariable theories. Phys. Rev. Lett. 23, 880 (1969).
Brunner, N., Cavalcanti, D., Pironio, S., Scarani, V. & Wehner, S. Bell nonlocality. Rev. Mod. Phys. 86, 419 (2014).
Hensen, B. et al. Loopholefree Bell inequality violation using electron spins separated by 1.3 kilometres. Nature 526, 682–686 (2015).
Giustina, M. et al. Significantloopholefree test of Bell’s theorem with entangled photons. Phys. Rev. Lett. 115, 250401 (2015).
Shalm, L. K. et al. Strong loopholefree test of local realism. Phys. Rev. Lett. 115, 250402 (2015).
Li, M.H. et al. Test of local realism into the past without detection and locality loopholes. Phys. Rev. Lett. 121, 080404 (2018).
Rosenfeld, W. et al. Eventready Bell test using entangled atoms simultaneously closing detection and locality loopholes. Phys. Rev. Lett. 119, 010402 (2017).
Arute, F. et al. Quantum supremacy using a programmable superconducting processor. Nature 574, 505–510 (2019).
Kurpiers, P. et al. Deterministic quantum state transfer and remote entanglement using microwave photons. Nature 558, 264–267 (2018).
Walter, T. et al. Rapid, highfidelity, singleshot dispersive readout of superconducting qubits. Phys. Rev. Appl. 7, 054020 (2017).
Magnard, P. et al. Microwave quantum link between superconducting circuits housed in spatially separated cryogenic systems. Phys. Rev. Lett. 125, 260502 (2020).
Scarani, V. Bell Nonlocality (Oxford Univ. Press, 2019).
Freedman, S. J. & Clauser, J. F. Experimental test of local hiddenvariable theories. Phys. Rev. Lett. 28, 938 (1972).
Aspect, A., Grangier, P. & Roger, G. Experimental realization of EinsteinPodolskyRosenBohm gedankenexperiment: a new violation of Bell’s inequalities. Phys. Rev. Lett. 49, 91 (1982).
Larsson, J.A. Loopholes in Bell inequality tests of local realism. J. Phys. A: Math. Theor. 47, 424003 (2014).
Aspect, A., Dalibard, J. & Roger, G. Experimental test of Bell’s inequalities using time varying analyzers. Phys. Rev. Lett. 49, 1804 (1982).
Weihs, G., Jennewein, T., Simon, C., Weinfurter, H. & Zeilinger, A. Violation of Bell’s inequality under strict Einstein locality conditions. Phys. Rev. Lett. 81, 5039 (1998).
Rowe, M. A. et al. Experimental violation of a Bell’s inequality with efficient detection. Nature 409, 791–794 (2001).
Aspect, A. Closing the door on Einstein and Bohr’s quantum debate. Physics 8, 123 (2015).
Acín, A. et al. Deviceindependent security of quantum cryptography against collective attacks. Phys. Rev. Lett. 98, 230501 (2007).
Šupić, I. & Bowles, J. Selftesting of quantum systems: a review. Quantum 4, 337 (2020).
Sekatski, P., Bancal, J.D., Wagner, S. & Sangouard, N. Certifying the building blocks of quantum computers from Bell’s theorem. Phys. Rev. Lett. 121, 180505 (2018).
Vazirani, U. & Vidick, T. Fully deviceindependent quantum key distribution. Phys. Rev. Lett. 113, 140501 (2014).
Colbeck, R. Quantum and Relativistic Protocols for Secure MultiParty Computation. PhD thesis, Univ. Cambridge (2009).
Pironio, S. et al. Random numbers certified by Bell’s theorem. Nature 464, 1021–1024 (2010).
Colbeck, R. & Renner, R. Free randomness can be amplified. Nat. Phys. 8, 450–453 (2012).
Kessler, M. & ArnonFriedman, R. Deviceindependent randomness amplification and privatization. IEEE J. Selected Areas Inf. Theory 1, 568–584 (2020).
Krinner, S. et al. Realizing repeated quantum error correction in a distancethree surface code. Nature 605, 669–674 (2022).
Leggett, A. J. Macroscopic quantum systems and the quantum theory of measurement. Progr. Theor. Phys. Suppl. 69, 80–100 (1980).
Clarke, J. & Wilhelm, F. K. Superconducting quantum bits. Nature 453, 1031–1042 (2008).
Martinis, J. M., Devoret, M. H. & Clarke, J. Energylevel quantization in the zerovoltage state of a currentbiased Josephson junction. Phys. Rev. Lett. 55, 1543 (1985).
Ansmann, M. et al. Violation of Bell’s inequality in Josephson phase qubits. Nature 461, 504–506 (2009).
The BIG Bell Test Collaboration. Challenging local realism with human choices. Nature 557, 212–216 (2018).
Zhong, Y. P. et al. Violating Bell’s inequality with remotely connected superconducting qubits. Nat. Phys. 15, 741–744 (2019).
Wootters, W. K. Entanglement of formation of an arbitrary state of two qubits. Phys. Rev. Lett. 80, 2245 (1998).
Horodecki, R., Horodecki, P. & Horodecki, M. Violating Bell inequality by mixed spin12 states: necessary and sufficient condition. Phys. Lett. A 200, 340 (1995).
Garg, A. & Mermin, N. D. Detector inefficiencies in the EinsteinPodolskyRosen experiment. Phys. Rev. D 35, 3831 (1987).
Eberhard, P. H. Background level and counter efficiencies required for a loopholefree EinsteinPodolskyRosen experiment. Phys. Rev. A 47, R747 (1993).
Magnard, P. MeterScale Microwave Quantum Networks for Superconducting Circuits. PhD thesis, ETH Zurich (2021).
Axline, C. et al. Ondemand quantum state transfer and entanglement between remote microwave cavity memories. Nat. Phys. 14, 705–710 (2018).
CampagneIbarcq, P. et al. Deterministic remote entanglement of superconducting circuits through microwave twophoton transitions. Phys. Rev. Lett. 120, 200501 (2018).
Kurpiers, P., Walter, T., Magnard, P., Salathe, Y. & Wallraff, A. Characterizing the attenuation of coaxial and rectangular microwavefrequency waveguides at cryogenic temperatures. EPJ Quant. Technol. 4, 8 (2017).
D’Addabbo, A. et al. The CUORE cryostat. J. Low Temp. Phys. 193, 867–875 (2018).
Pechal, M. et al. Microwavecontrolled generation of shaped single photons in circuit quantum electrodynamics. Phys. Rev. X 4, 041010 (2014).
Abellán, C., Amaya, W., Mitrani, D., Pruneri, V. & Mitchell, M. W. Generation of fresh and pure random numbers for loopholefree Bell tests. Phys. Rev. Lett. 115, 250403 (2015).
Burkhart, L. D. et al. Errordetected state transfer and entanglement in a superconducting quantum network. PRX Quantum 2, 030321 (2021).
Barrett, J., Hardy, L. & Kent, A. No signalling and quantum key distribution. Phys. Rev. Lett. 95, 010503 (2005).
Peñas, G. F., Puebla, R., Ramos, T., Rabl, P. & GarcíaRipoll, J. J. Universal deterministic quantum operations in microwave quantum links. Phys. Rev. Applied 17, 054038 (2022).
Kjaergaard, M. et al. Superconducting qubits: current state of play. Ann. Rev. Condens. Matter Phys. 11, 369–395 (2020).
Josephson, B. D. Possible new effects in superconductive tunnelling. Phys. Lett. 1, 251–253 (1962).
Koch, J. et al. Chargeinsensitive qubit design derived from the Cooper pair box. Phys. Rev. A 76, 042319 (2007).
DiCarlo, L. et al. Demonstration of twoqubit algorithms with a superconducting quantum processor. Nature 460, 240–244 (2009).
Blais, A., Grimsmo, A. L., Girvin, S. M. & Wallraff, A. Circuit quantum electrodynamics. Rev. Mod. Phys. 93, 025005 (2021).
Cirac, J. I., Zoller, P., Kimble, H. J. & Mabuchi, H. Quantum state transfer and entanglement distribution among distant nodes in a quantum network. Phys. Rev. Lett. 78, 3221 (1997).
Scheidl, T. et al. Violation of local realism with freedom of choice. Proc. Natl Acad. Sci. USA 107, 19708 (2010).
Zhang, Y., Glancy, S. & Knill, E. Asymptotically optimal data analysis for rejecting local realism. Phys. Rev. A 84, 062118 (2011).
Acknowledgements
We thank M. Frey, N. Kohli, R. Schlatter, A. Fauquex, R. Keller, B. Dönmez, M. Hinderling, S. Wili, F. Marxer, A. Schwarzer and J.A. Agner for technical support on setting up the laboratory, and designing, building and testing the cryogenic system. We thank J. Herrmann, L. Raabe, N. Mostaan, E. Portolés, M. Ruckriegel and J. Heinsoo for their contributions to software and electronics. We thank A. Aspect, G. Blatter, N. Gisin, R. Hanson, A. Imamoglu, R. Renner and R. Wolf for commenting on an early version of the manuscript. The work at ETH Zurich was funded by the European Research Council through the ‘Superconducting Quantum Networks’ (SuperQuNet) project, by the European Union’s Horizon 2020 FETOpen project SuperQuLAN (grant no. 899354), by the National Centre of Competence in Research ‘Quantum Science and Technology’ (NCCR QSIT), a research instrument of the Swiss National Science Foundation, and by ETH Zurich. B.R. and A.B. acknowledge support from the Natural Sciences and Engineering Research Council of Canada, the Canada First Research Excellence Fund and from the Vanier Canada Graduate Scholarships. J.D.B. and N.S. acknowledge support by the Institut de Physique Théorique, Commissariat à l’Energie Atomique et aux Energies Alternatives, by the European HighPerformance Computing Joint Undertaking under grant agreement no. 101018180 and project name HPCQS and by a French national quantum initiative managed by Agence Nationale de la Recherche in the framework of France 2030 with the references ANR22PETQ0007, project name EPIQ and ANR22PETQ0009, project name DIQKD. M.W.M. acknowledges support by NextGenerationEU (grant no. PRTRC17.I1) and by projects SAPONARIA (grant no. PID2021123813NBI00) and MARICHAS (grant no. PID2021126059OAI00), by ‘Severo Ochoa’ Center of Excellence CEX2019000910S, Generalitat de Catalunya through the CERCA program, AGAUR grant no. 2021SGR01453, by Fundació Privada Cellex and by Fundació MirPuig.
Funding
Open access funding provided by Swiss Federal Institute of Technology Zurich.
Author information
Authors and Affiliations
Contributions
S.S., J.S., A.K. and P.M. planned and performed the experiment and analysed the data. S.S., P.M., T.W. and J.S. designed and tested the quantum devices. J.S., J.L., S.S., P.M., P.K. and A.W. devised and tested the cryogenic setup. J.C.B., M.G., G.J.N., T.W. and A.R. fabricated the devices. P.M. and S.S. integrated the measurement basis choice scheme. A.K., A.C., S.S., J.S. and P.M. developed and implemented the setup synchronization scheme and the verification procedure for closing the locality loophole. S.S., A.C., K.R., P.M. and A.A. developed the experiment control and FPGAbased data analysis code. B.R., S.S. and A.B. ran the master equation simulations. F.M., J.M., W.A., M.W.M., C.A. and P.M. developed and tested the RNGs. J.D.B. and N.S. assisted with the statistical analysis and the manuscript. S.S., J.S. and P.M. provided the figures for the manuscript. S.S. and A.W. wrote the manuscript with major contributions from A.K., J.S., J.D.B., N.S. and M.W.M. and input from all authors. A.W. supervised the project.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature thanks Peter Bierhorst, Yanbao Zhang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Supplementary Information
This Supplementary Information file contains 11 sections, Supplementary Figs. 1–12, Tables 1–6 and additional references.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Storz, S., Schär, J., Kulikov, A. et al. Loopholefree Bell inequality violation with superconducting circuits. Nature 617, 265–270 (2023). https://doi.org/10.1038/s41586023058850
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41586023058850
This article is cited by

My Discussions of Quantum Foundations with John Stewart Bell
Foundations of Science (2024)

A Formalization of the CHSH Inequality and Tsirelson’s Upperbound in Isabelle/HOL
Journal of Automated Reasoning (2024)

Electric fieldtuneable crossing of hole Zeeman splitting and orbital gaps in compressively strained germanium semiconductor on silicon
Communications Materials (2023)

Nonlocality with massive particles and mesoscopic systems
Nature Reviews Physics (2023)

Microwavetooptical conversion in a roomtemperature 87Rb vapor for frequencydivision multiplexing control
Communications Physics (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.