Abstract
Quantum technologies promise advantages over their classical counterparts in the fields of computation, security and sensing. It is thus desirable that classical users are able to obtain guarantees on quantum devices, even without any knowledge of their inner workings. That such classical certification is possible at all is remarkable: it is a consequence of the violation of Bell inequalities by entangled quantum systems. Deviceindependent selftesting refers to the most complete such certification: it enables a classical user to uniquely identify the quantum state shared by uncharacterized devices by simply inspecting the correlations of measurement outcomes. Selftesting was first demonstrated for the singlet state and a few other examples of selftestable states were reported in recent years. Here, we address the longstanding open question of whether every pure bipartite entangled state is selftestable. We answer it affirmatively by providing explicit selftesting correlations for all such states.
Introduction
Since it was proposed a decade ago^{1}, the deviceindependent certification of quantum devices has attracted a lot of interest because it requires minimal assumptions: the nosignalling constraint on the devices, and the validity of quantum theory. If these are accepted, the certification can then be performed by a purely classical user, that queries the devices with classical inputs and observes the correlations in the classical outputs. This is possible thanks to the violation of Bell inequalities^{2}, and constitutes the operational interpretation of this phenomenon.
In a deviceindependent way, one can bound specific quantities like the amount of randomness^{3}, the length of the secret key in quantum cryptography^{1} or the dimension of the Hilbert space of the systems involved^{4}. But for some correlations, the characterization can be as complete as one can hope for. Indeed, certain correlations can be achieved exclusively by measurements on a unique quantum state (up to local transformations). We adopt the technical term ‘deviceindependent selftesting’ to refer to such a certification. Selftesting correlations can be thought of as a classical fingerprint of a state.
The fact that a purely classical user can certify the quantum state of a system is in contrast with the usual quantum state tomography, which relies on the characterization of the degrees of freedom under study and the corresponding measurements. In this case, a classical user lacking knowledge of the inner workings of a quantum device would have no choice but to trust that it has been manufactured according to specifications.
The most celebrated example of a state that can be selftested is the maximally entangled pair of qubits (the ‘singlet’ state). One selftesting criterion for this state is the maximal violation of the wellknown ClauserHorneShimonyHolt (CHSH) inequality^{5,6}. Another criterion was put forward by Mayers and Yao, in the paper which coined the term ‘selftesting’^{7}. Since then, selftesting of the twoqubit singlet has been made robust^{8}, extended to sequential^{9} and parallel certification of many copies^{10,11,12,13,14,15}, and its complete set of selftesting criteria with two dichotomic measurements has been provided^{16}. A variety of other quantum states have also been proved to be selftestable: all partially entangled pure twoqubit states^{17,18}, the maximally entangled pair of qutrits^{19}, the partially entangled pair of qutrits that violates maximally the CGLMP_{3} inequality^{20,21} and a small class of higherdimensional partially entangled pairs of qudits, through results in parallel selftesting^{12}. For the multipartite case, selftesting is known for the family of graph states^{22,23} and for a few nongraph threequbit states^{23,24}. Hence, it is clear that selftesting is not an exclusive characteristic of maximally entangled states nor qubit states. However, little is known about selftesting of higherdimensional entangled states (that is, states of entangled qudits for d>2).
In this work, we prove that all pure bipartite entangled quantum states can be selftested, by constructing explicit correlations built on the framework outlined by Yang and Navascués^{17}.
Results
The 3d4d Bell scenario
We work in a bipartite Bell scenario, and we refer to Alice and Bob as operating the uncharacterized devices (or rather as the devices themselves). They receive inputs x and y, respectively, from the classical verifier, corresponding to their choice of measurement settings, and they return outcomes a and b respectively. In the particular scenario that we will consider, Alice has three possible measurement settings and Bob has four, while they have d possible outcomes each. So the inputs are x∈{0, 1, 2} and y∈{0, 1, 2, 3} and the outputs are a, b∈{0, 1, 2, ⋯, d−1}. We refer to this as a [{3, d},{4, d}] Bell scenario (Fig. 1a). The result of this Bell experiment can be fully described by the probabilities P(a, bx, y) of obtaining a pair of outcomes a, b on measurement settings x, y. In the deviceindependent approach, the dimensionality of the measured system is not bounded a priori. Hence, the measurements made on the system can be assumed to be projective, with the projection corresponding to Alice obtaining outcome a on measurement setting x, and likewise for on Bob’s side. No further characterization of either the state or the measurements is required, and estimating the P(a, bx, y) is all that has to be done in the lab.
Selftesting of all pure bipartite entangled states
We state our main theorem.
Theorem 1: for every bipartite entangled state of qudits , there exist [{3, d},{4, d}] quantum correlations that, when reproduced by Alice and Bob through local measurements on a joint state ρ, imply the existence of a local isometry Φ such that , where is some auxiliary state. Moreover, under the isometry Φ, the local measurements on ρ are equivalent to measurements that act trivially on and as the ideal measurements on (described exactly in the Supplementary Methods).
The proof of Theorem 1 now proceeds at the mathematical level (Fig. 1c), and we provide an overview of the main ideas. The full details are contained in the Supplementary Methods. For ease of exposition, we take Alice and Bob’s shared state to be a pure state , but our proof goes through in the same way for a general ρ. Initially, the verifier has no knowledge about the state shared by the two devices, and he wishes to certify that it is a specific state of two qudits. We can think of providing Alice and Bob with a qudit each (A′ and B′), initialized in an arbitrary state ; then trying to swap information from the two blackboxes into these qudits. If at the end of the swap one finds , one concludes that the boxes contained the state ⊗ before the swap, where the precise state is not important, and is just ancillary. The physical and mathematical parts of selftesting are connected by the existence of a swap operation, which acts as desired thanks to the constraints given by the P(a, bx, y). In mathematical terms, what we have just explained amounts to constructing a local isometry Φ such that . If such an isometry exists, one says that these correlations selftest . Invoking the Schmidt decomposition, selftesting all bipartite entangled states reduces to selftesting all states of the form.
where 0<c_{i}<1 for all i and .
One may wonder whether mixed states could also be selftested, that is, if some P(a, bx, y) is uniquely compatible with a mixed state (or with its purified version, but with measurements acting trivially on the purifying system). The answer is negative: any P(a, bx, y) produced by a bipartite mixed state can be reproduced by a bipartite pure state of the same dimension^{25}. Hence, in the bipartite scenario, the best one can hope for is to selftest every pure state. To illustrate how we construct selftesting correlations for such a target state as in equation (1), we look at the case d=4, so that . We already know that with correlations having two inputs per party, one can selftest any twoqubit state (that is, d=2)^{17,18}. So, the idea is that for x, y∈{0, 1}, we choose P(a, bx, y) so that the probabilities for a, b∈{0, 1} certify , while those for a, b∈{2, 3} certify . All the other P(a, bx, y), that is, those where (a, b){0, 1}^{2}∪{2, 3}^{2}, are set to zero. Then, one similarly uses measurement settings x∈{0, 2} and y∈{2, 3}, but with a block structure certifying and .
In other words, our correlations rely on detecting a pattern of twoqubit correlations compatible exclusively with , across a suitable directsum decomposition of the Hilbert space in which the joint state lies. The recipe is clearly not restricted to d=4: with the same number of measurement settings, and naturally generalized blockdiagonal correlations, one can selftest any bipartite entangled pure state of any dimension (see Fig. 2 for an illustration for d even; the argument carries on to d odd as well).
Proof outline of Theorem 1
While the recipe is intuitive, the formal proof must follow the scheme illustrated in Fig. 1, and thus construct the local isometry. All the technical details are given in the Supplementary Methods, and here we outline how the proof proceeds.
First, we need to formalize the intuition that the twoqubit blocks are certified by the blockdiagonal correlations described earlier. Consider the ‘tilted CHSH’ Belltype inequality^{26}
where x, y, a, b∈{0, 1}, α∈[0, 2), and . It is known, thanks to Yang and Navascués^{17}, and Bamps and Pironio^{18}, that maximal violation of this inequality, corresponding to , selftests the state , with . However, when we try to apply this certification to each consecutive pair of two outcomes, we find that the value of the left hand side (LHS) of inequality (2) in each block, computed from the P(a, bx, y) we described earlier, is the maximal violation multiplied by the probabilistic weight of that block: in other words, it is not the maximal violation itself. To recognize the covert maximal violation that indeed resides in each block, and the certification that follows from it, one has to realize that the state which achieves the maximal violation is not the joint state , but rather its projection onto each 2 × 2 block. From each such maximal violation, one can construct the four operators , , with support on the (2m, 2m+1) block (or , with support on the (2m+1, 2m+2) block), that are used in the selftesting isometry from Yang and Navascués^{17}, and Bamps and Pironio^{18}.
Second, one has to tie together the certifications in the different blocks, and explicitly construct the overall local isometry Φ such that . A sufficient condition for the existence of such an isometry has been formulated by Yang and Navascués^{17}: one needs complete sets of orthogonal projections and and unitary operators satisfying the following conditions for all k=0, 1, ..., d−1:
where ω=e^{2πi/d}. In our construction, the projections are chosen from Alice and Bob’s projection measurements, and each operator is the product of all the and (formally extended to the whole space, and denoted and respectively in the Supplementary Methods) covering all 2 × 2 blocks up to k. This product spans the alternating block structure: it is in these operators that the crucial connection between blocks is encoded. It is not difficult, finally, to extend the proof of selftesting to the ideal measurements (see the Supplementary Methods).
Discussion
In conclusion, we have proved the longstanding conjecture that all bipartite entangled quantum states can be selftested, by explicitly providing a ‘classical fingerprint’, or selftesting correlations, for every such state. Such fingerprints are not unique: our proof also remains valid if, in each block, the criterion based on the tilted CHSH inequality is replaced by any other criterion that selftests the same twoqubit state. In particular, through the correlations adopted in Yang and Navascués^{17}, a maximally entangled pair of qudits can be selftested with only three measurements per side, that is, in the[{3, d},{3, d}] Bell scenario. We have only presented the proof of ideal selftesting (when the correlations are exact): while we believe that some robustness bounds can be derived, existing analytical tools produce notoriously unsatisfying bounds, and the numerical tools that give much better bounds can only be applied to selected examples. In this situation, we would rather wait for progress in analytical tools, of the kind shown by Kaniewski^{27}.
Besides shedding new light on quantum states and quantum correlations, our result has potential applications to quantum technologies. Proofs of certification of quantum devices, from randomness to cryptography and ultimately quantum computing, have often been based on a selftesting criterion, the rigidity of the CHSH game^{9,28,29}. Our work adds total flexibility of choosing the state in the bipartite scenario. One direct application may be in the context of quantum random number generation. Concretely, in deviceindependent randomness expansion (the first deviceindependent randomnumber generation scheme to be proposed, and the only to have been experimentally implemented to date^{3}), guaranteed private randomness is generated from an initial random seed. Based on our selftesting procedure, a small random seed (two random trits) could provide up to O(log d) bits of private randomness per run, with d limited only by the experimental stateoftheart. Indeed, in the ideal case, if one knows that the global state is maximally entangled, each outcome of any ideal local measurement has probability 1/d. A robustness analysis for selftesting both the state and the measurements is required to assess the expansion rate of any protocol based on our selftesting procedure, and we leave this for future work. Any such protocol would become feasible as soon as one can realize loopholefree Bell tests with entangled states of dimension d.
Data availability
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Additional information
How to cite this article: Coladangelo, A. et al. All pure bipartite entangled states can be selftested. Nat. Commun. 8, 15485 doi: 10.1038/ncomms15485 (2017).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
 1
Acín, A. et al. Deviceindependent security of quantum cryptography against collective attacks. Phys. Rev. Lett. 98, 230501 (2007).
 2
Bell, J. S. On the einstein podolsky rosen paradox. Physics 1, 195–200 (1964).
 3
Pironio, S. et al. Random numbers certified by Bell’s theorem. Nature 464, 1021–1024 (2010).
 4
Brunner, N. et al. Testing the Dimension of Hilbert Spaces. Phys. Rev. Lett. 100, 210503 (2008).
 5
Summers, S. J. & Werner, R. Bell’s inequalities and quantum field theory. I. General setting. J. Math. Phys. 28, 2448–2456 (1987).
 6
Popescu, S. & Rohrlich, D. Which states violate Bell’s inequality maximally? Phys. Lett. A 169, 411–414 (1992).
 7
Mayers, D. & Yao, A. Selftesting quantum apparatus. Quantum Inf. Comput. 4, 273–286 (2004).
 8
McKague, M., Yang, T. H. & Scarani, V. Robust self testing of the singlet. J. Phys. A Math. Theor. 45, 455304 (2012).
 9
Reichardt, B. W., Unger, F. & Vazirani, U. Classical command of quantum systems. Nature 496, 456–460 (2013).
 10
Wu, X., Bancal, J.D., McKague, M. & Scarani, V. Deviceindependent parallel selftesting of two singlets. Phys. Rev. A 93, 062121 (2016).
 11
McKague, M. Selftesting in parallel. New J. Phys. 18, 045013 (2016).
 12
Coladangelo, A. W. Parallel selftesting of (tilted) EPR pairs via copies of (tilted) CHSH. Preprint at https://arxiv.org/abs/1609.03687 (2016).
 13
Coudron, M. & Natarajan, A. The parallelrepeated magic square game is rigid. Preprint at https://arxiv.org/abs/1609.06306 (2016).
 14
Chao, R., Reichardt, B. W., Sutherland, C. & Vidick, T. Test for a large amount of entanglement, using few measurements. Preprint at https://arxiv.org/abs/1610.00771 (2016).
 15
Natarajan, A. & Vidick, T. Robust selftesting of manyqubit states. Preprint at https://arxiv.org/abs/1610.03574 (2016).
 16
Wang, Y., Wu, X. & Scarani, V. All the selftestings of the singlet for two binary measurements. New J. Phys. 18, 025021 (2016).
 17
Yang, T. H. & Navascués, M. Robust selftesting of unknown quantum systems into any entangled twoqubit states. Phys. Rev. A 87, 050102 (2013).
 18
Bamps, C. & Pironio, S. Sumofsquares decompositions for a family of ClauserHorneShimonyHoltlike inequalities and their application to selftesting. Phys. Rev. A 91, 052111 (2015).
 19
Salavrakos, A. et al. Bell inequalities for maximally entangled states. Preprint at https://arxiv.org/abs/1607.04578 (2016).
 20
Yang, T. H., Vértesi, T., Bancal, J.D., Scarani, V. & Navascués, M. Robust and versatile blackbox certification of quantum devices. Phys. Rev. Lett. 113, 040401 (2014).
 21
Acín, A., Durt, T., Gisin, N. & Latorre, J. I. Quantum nonlocality in two threelevel systems. Phys. Rev. A 65, 052325 (2002).
 22
McKague, M. Selftesting graph states. Conference on Quantum Computation, Communication, and Cryptography 104–120Springer (2011) https://link.springer.com/book/10.1007/9783642544293.
 23
Pál, K. F., Vértesi, T. & Navascués, M. Deviceindependent tomography of multipartite quantum states. Phys. Rev. A 90, 042340 (2014).
 24
Wu, X. et al. Robust selftesting of the threequbit w state. Phys. Rev. A 90, 042339 (2014).
 25
Sikora, J., Varvitsiotis, A. & Wei, Z. Minimum dimension of a hilbert space needed to generate a quantum correlation. Phys. Rev. Lett. 117, 060401 (2016).
 26
Acín, A., Massar, S. & Pironio, S. Randomness versus nonlocality and entanglement. Phys. Rev. Lett. 108, 100402 (2012).
 27
Kaniewski, J. Analytic and nearly optimal selftesting bounds for the clauserhorneshimonyholt and mermin inequalities. Phys. Rev. Lett. 117, 070402 (2016).
 28
Coudron, M. & Yuen, H. Infinite randomness expansion with a constant number of devices. in STOC ‘14 Proceedings of the fortysixth annual ACM symposium on Theory of computing 427–436ACM (2014) http://dl.acm.org/citation.cfm?id=2591796.
 29
Miller, C. A. & Shi, Y. Robust protocols for securely expanding randomness and distributing keys using untrusted quantum devices. J. ACM 63, 33 (2016).
Acknowledgements
We thank Matthew McKague and Thomas Vidick for comments on earlier drafts, and acknowledge discussions with them as well as with Miguel Navascués and Xingyao Wu. This research is supported by the Singapore Ministry of Education Academic Research Fund Tier 3 (Grant No. MOE2012T31009); by the National Research Fund and the Ministry of Education, Singapore, under the Research Centres of Excellence programme. A.C. is supported by AFOSR YIP award number FA95501610495.
Author information
Affiliations
Contributions
All the authors contributed to all aspects of this work.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Supplementary Information
Supplementary tables, supplementary methods and supplementary references. (PDF 332 kb)
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
Coladangelo, A., Goh, K. & Scarani, V. All pure bipartite entangled states can be selftested. Nat Commun 8, 15485 (2017). https://doi.org/10.1038/ncomms15485
Received:
Accepted:
Published:
Further reading

SelfTesting of Physical Theories, or, Is Quantum Theory Optimal with Respect to Some InformationProcessing Task?
Physical Review Letters (2020)

Experimental Optimal Verification of Entangled States Using Local Measurements
Physical Review Letters (2020)

Selftesting of quantum systems: a review
Quantum (2020)

A generalization of CHSH and the algebraic structure of optimal strategies
Quantum (2020)

Simultaneous certification of entangled states and measurements in bounded dimensional semiquantum games
Physical Review Research (2020)
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.